international journal of computing and ict...

International Journal of Computing and ICT Research, Vol. 11, Issue 1, June 2017

1

Contents Volume 11, No 1, June 2017.

ISSN 1818-1139 (PRINT), ISSN 1996-1065 (ONLINE)

Is Africa “Overproducing” “Well Trained” Information Technologists? – Part – 1 …………………… 6 Joseph M. Kizza – Editor-in-Chief Is the SAMR Model Valid and Reliable for Measuring the Use of ICT in Pedagogy? Answers from a Study of Teachers of Mathematical Disciplines in Universities in Uganda………11 Marjorie S K Batiibwe; Fred E K Bakkabulindi; John M Mango Empirical Study of Continuous Change of Open Source System…………………..………………………….31 Michael Abayomi Olatunji, Rufus Olalere Oladele and Amos Orenyi Bajeh Measuring the Impacts of E-Learning on Students’ Achievement in Learning Process: An Experience from Tanzanian Public Universities……………………………………………………………………….53 Titus Tossy Reducing Checkpoint Overhead in Grid Environment……………………………………………………………..72 Faki Ageebe Silas, Jimoh Rasheed Gbenga Keystroke Dynamics Authentication for a Web-Based Sales and Stock Solution……………………..84 Oluwakemi C. Abikoye and Bilikis T. Sanni

International Journal

of Computing and

ICT Research

2 International Journal of Computing and ICT Research, Vol. 11, Issue 1, June 2017

International Journal of Computing and ICT Research

Editorial Board

Editor-in-Chief: Prof. Joseph M. Kizza,

Department of Computer Science and Engineering

College of Engineering and Computer Science

The University of Tennessee-Chattanooga,

615 McCallie Avenue, Chattanooga, Tennessee, USA

[email protected]

Managing Editors:

Information Technology

Prof. Shushma Patel, London South Bank University, UK

Information Systems

Prof. Ravi Nath, Creighton University, Nebraska, USA

Software Engineering

Prof. P.K. Mahanti, University of New Brunswick, Canada

Data Communication and Computer Networks

Prof. Vir Phoha, Syracuse University, New York , USA

Production Editor:

Journal Editorial Office:

The International Journal of Computing and ICT Research

Makerere University

P.O. Box 7062,

Kampala, Uganda.

Tel: +256 414 540628

Fax: +256 414 540620

Email: [email protected]

Web: http://www.ijcir.mak.ac.ug

mailto:[email protected]

http://www.ijcir.mak.ac.ug/


3

Volume 11, Issue 1 June 2017

The International Journal of Computing and ICT Research

College of Computing and Information Sciences

Makerere University

P.O. Box 7062,

Kampala, Uganda.

Tel: +256 414 540628

Fax: +256 414 540628

Email: [email protected]

International Journal of Computing

and ICT Research



Web: http://www.ijcir.mak.ac.ug

Book Reviews

Every issue of the journal will carry one or more book reviews. This is a call for reviewers of

books. The book reviewed must be of interest to the readers of the journal. That is to say, the book

must be within the areas the journal covers. The reviews must be no more than 500 words. Send

your review electronically to the book review editor at: [email protected]



5

International Journal of Computing and ICT Research

The IJCIR is an independent biannual publication of Makerere University. In addition to

publishing original work from international scholars from across the globe, the Journal strives to

publish African original work of the highest quality that embraces basic information

communication technology (ICT) that will not only offer significant contributions to scientific

research and development but also take into account local development contexts. The Journal

publishes papers in computer science, computer engineering, software engineering, information

systems, data communications and computer networks, ICT for sustainable development, and

other related areas. Two issues are published per year: June and December. For more detailed

topics please see: http://www.ijcir.mak.ac.ug

Submitted work should be original and unpublished current research in computing and ICT based

on either theoretical or methodological aspects, as well as various applications in real world

problems from science, technology, business or commerce.

Short and quality articles (not exceeding 20 single spaced type pages) including references are

preferable. The selection of journal papers which involves a rigorous review process secures the

most scholarly, critical, analytical, original, and informative papers. Papers are typically published

in less than half a year from the time a final corrected version of the manuscript is received.

Authors should submit their manuscripts in Word or PDF to [email protected]. Manuscripts

submitted will be admitted subject to adherence to the publication requirements in formatting and

style. For more details on manuscript formatting and style please visit the journal website at:

http://www.ijcir.mak.ac.ug.





Is Africa “Overproducing” “Well Trained” Information Technologists?

– Part 1

PROF. JOSEPH M. KIZZA1,

Editor-in-Chief

Department of Computer Science and Engineering,

The University of Tennessee-Chattanooga, Tennessee, 37403, USA.

IJCIR Reference Format:

Kizza, Joseph. M. Is Africa “Overproducing” “Well Trained” Information Technologists? – Part

1, Vol. 11, Issue.1 pp 6 - 10. http://ijcir.mak.ac.ug/volume11-issue1/article1.pdf

INTRODUCTION

In my editorial article of the maiden issue, Vol. 1 Issue 1, Vol. 3 Issue 2 and in subsequent volumes

I pointed out that the African technology journey from the bottom of the stack to where Africa is

now, has not been without problems. There, at the dawn of the Information and Communications

Technology (ICT), Africa’s technological imprints were but non-starters compared to the giant

steps that were being taken in the rest of the world. Throughout the continent, ICT capacity and

infrastructure were low and in some places non-existent, and equipment acquisition was sporadic

and unplanned. However, as I pointed out then and stress now, Africans were just deprived not

disabled. African leading universities and institutions set themselves on a quest to be the ICT

incubators and jump start ICT education and research and build ICT capacity to help in the

construction of the badly needed infrastructure.

1 Author’s Address: Joseph M. Kizza, Department of Computer Science and Engineering,The University of Tennessee-

Chattanooga,Chattanooga, TN 37403, USA., [email protected]. "Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are

not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for

components of this work owned by others than IJCIR must be honored. Abstracting with credit is permitted. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee."

© International Journal of Computing and ICT Research 2008.

International Journal of Computing and ICT Research, ISSN 1818-1139 (Print), ISSN 1996-1065 (Online), Vol.11, Issue 1, pp. 6 - 10, June 2017.

http://ijcir.mak.ac.ug/volume11-issue1/article1.pdf


7

For these institutions and everybody else involved in the quest for technological advance, the climb

to the top of the technological mountain has been and continues to be steep. However, with the

typical African determination, they are inching on to the top. Years into the climb, there has been

signs of achievements. Everywhere in Africa today, one witnesses exuberance, soaring interest,

especially of the young, an increasing inventory of working ICT equipment, an increasing number

of confident and young ICT technocrats, an unbelievably large number of young people taking

courses in information technology and a growing number of governments believing more in ICT

as a tool for development. The African technological acquisition, though still low by international

standards, driven mostly by an unprecedented indigenous interest in technological development

and the numerous and sometimes ambitious initiatives by NGOs and the donor community, is

changing the fortunes of Africa, quickly leapfrogging her into the 21st

Century with rapid

development not seen in generations. The long awaited African technological dawn may be in

sight.

But like all good and successful journeys, it is time to take a break and a breather and take stock

of what is it that we are getting. Indeed the results we are getting, though still overwhelmingly

good, there are signs that things are not going as planned. There are signs of cracks in the castles

of success we are building. The first and probably the biggest crack to show is the growing numbers

of university graduates in ICT that are roaming African streets looking for employment. This

should lead us to serious soul searching for where we went wrong. Is Africa

“Overproducing” “Well Trained” Information Technologists?

Why do we have what seems to be so many frustrated young ICT graduated that are roaming our

streets with annoyed parents who spent their lives savings for a promise that was not to be? Did

we oversell ICT? Did we false advertise ICT? Did we train their sons and daughters right? Have

their governments failed to create enabling environments and infrastructures to enable these

graduates to start on their own? Last and most important, did we misunderstand ICT? To try to

understand what is going on and ignite a debate about this very growing problem, I put quotes of

key words and phrases in my title question to indicate how loaded the question is.


As we start to debate the question looking for answers, I want to point out pertinent issues that

may explain some of the root courses. I discussed these very same issues, on the cause of poor

education given to our students, in Vol. 5 Issue 2:

A PERVASIVE CONSULTANCY CULTURE Today, intellectual life in universities has

been reduced to bare-bones classroom activity. Extra-curricular seminars and workshops

have migrated to hotels. Workshop attendance goes with transport allowances and per

diem. All this is part of a larger process, the NGO-ization of the university. Academic

papers have turned into corporate-style power point presentations. Academics read less and

less. A chorus of buzz words have taken the place of lively debates (Mamdani, Mail and

Guardian Online)

Mahmood Mamdani: “African Universities Breed ―Native Informers‖, Not Researchers”

- A leading East African political scientist, Prof. Mahmood Mamdani, who is the director

of Makerere University ‘s Institute of Social Research has put universities in Sub-Saharan

Africa in the dock by accusing them of not creating researchers but churning out native

informers to national and international non-governmental organizations. (Mamdani, Mail

and Guardian Online)

Statistics from the United Nations Educational Scientific and Cultural Organization

(UNESCO) reveal that the entire African continent contributes only 2.3 per cent of the

world ‘s researchers. (Wachira Kigotho, online)

UNESCO estimates that on average, Africa has only 169 researchers per one million

inhabitants. Apart from having the lowest density of researchers in the world, investment

in research and development in Africa stands at 0.9 per cent. (Stepps in Sync.)

Besides underrated education given to our students as one of the root causes of this problem, there

are other causes including:

Lack of enough resources like private capital funding and back loans to enable these

graduates to start on their own- building start ups. The majority of technology-based

businesses and companies start as individual start-ups.

African governments are not investing enough into policies and financing of ICT-based

enabling environments. There are a few exceptions like Kenya’s Konza Techno City, an


9

IT-focused “smart” metropolis – code named the Silicon Savannah (Jonathan Rosen).

False advertising by institutions of higher learning. In the last 10 years or so more

universities and institutions of higher learning have sprung up across African than in the

last fifty years. The key getter is ICT. Everyone of these “new” institutions, some not

worthy to be called institutions of higher learning, and with poorly trained and poorly paid

teaching faculty, are advertising to parents as they offer a strong curriculum of ICT and

how their sons and daughter will not have problems finding ICT-related jobs. Parents pull

their money out of mattresses to pay for the success of their off-springs.

Probably the most serious problem is the misunderstanding of what ICT is. I attended a

conference where this one African minister tried to convince everyone that “with “ICTs”

(whatever ITCs meant to him) we can do anything”. This has been and continues to be a

very serious and indeed dangerous problem. ICT is a spectrum of technologies that if one

is not careful, one can take courses and courses and come out without an iota of knowledge

of even writing a single line of code to do anything useful. I have seen African students

with degrees in ICT coming to American universities and not be able to pass a single

undergraduate freshman course in computing sciences. It is a deplorable situation we are

in. In this situation, how to you expect to graduate a student who can find a meaningful

job?

Something needs to give!

REFERENCES

MAHMOUD MAMDANI. The Importance of Research in a University - Mahmoud Mamdani,

http://pascalobservatory.org/pascalnow/blogentry/importance-research-university-mahmoud-

mamdani

WACHIRA KIGOTHO, Africa, Home to Only 2.3 per cent World ‘s Researchers,

http://www.standardmedia.co.ke/InsidePage.php?id=2000040499&cid=4&ttl=Africa+home+to+

only+2.3+per+cent+world%E2%80%99s+researchers

MAHMOUD MAMDANI, Africa ‘s Post-Colonial Scourge, Mail and Guardian Online.

http://www.ru.ac.za/media/rhodesuniversity/content/documents/institutionalplanning/African%2


0higher%20education%20-%20Mahmood%20Mamdani%2027May2011.pdf

STEPPS IN SYNC. About the Life of Creatives in the Overlooked ‗Pockets ‘of the Planet.

http://steppesinsync.wordpress.com/about-steppes-in-sync/

ROSEN, JONATHAN. Business Report. “Kenya Tries to Build Its Silicon Valley”.

https://www.technologyreview.com/s/543406/kenya-tries-to-build-its-silicon-valley/


11

Is the SAMR Model Valid and Reliable for Measuring the Use of ICT in

Pedagogy? Answers from a Study of Teachers of Mathematical

Disciplines in Universities in Uganda

*MARJORIE S K BATIIBWE2; *FRED E K BAKKABULINDI; **JOHN M MANGO

*College of Education & External Studies, Makerere University

**College of Natural Sciences, Makerere University

ABSTRACT

Puentedura's (2006) SAMR model of the use of ICT operationalises ICT as consisting of

substitution (S), augmentation (A), modification (M) and redefinition (R) constructs. Accordingly,

this study sought first, to establish the validity and reliability of each of the four constructs (S, A,

M & R). Second, to test whether the four constructs were independent. Third, to re-examine

whether the four-factor SAMR model of ICT was reasonable. A sample of 261 was chosen from

among teachers of mathematical disciplines in four universities in Uganda who reacted to a self-

administered questionnaire. The analysis involved using confirmatory factor analysis (CFA) and

Cronbach alpha (first objective); Pearson’s linear correlation (second objective); and exploratory

factor analysis (EFA) (third objective). CFA suggested that while not all the items of each of S

and A constructs were valid measures, all the items of each of M and R constructs were valid. The

four constructs were highly inter-correlated. EFA revealed that the four-factor SAMR model of

2 Author’s Address: MARJORIE S K BATIIBWE2; *FRED E K BAKKABULINDI; **JOHN M MANGO

*College of Education & External Studies, Makerere University, **College of Natural Sciences, Makerere University

"Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for

components of this work owned by others than IJCIR must be honored. Abstracting with credit is permitted. To copy otherwise, to republish, to

post on servers or to redistribute to lists, requires prior specific permission and/or a fee." © International Journal of Computing and ICT Research 2008.

International Journal of Computing and ICT Research, ISSN 1818-1139 (Print), ISSN 1996-1065 (Online), Vol.11, Issue 2, pp. 11 - 30, June

2017.


operationalising ICT was questionable. Hence a call was made to researchers to continue testing

the SAMR model in different contexts with intent to refining it.

General terms: Academic Staff, Cronbach Alpha, Factor Analysis, Mathematical disciplines, Use

of ICT in pedagogy


MARJORIE S K BATIIBWE; FRED E K BAKKABULINDI; JOHN M MANGO. Is the

SAMR Model Valid and Reliable for Measuring the Use of ICT in Pedagogy? Answers from a

Study of Teachers of Mathematical Disciplines in Universities in Uganda, Vol. 11, Issue.1 pp 11

- 30. http://ijcir.mak.ac.ug/volume11-issue1/article2.pdf

1. INTRODUCTION

Lloyd (2005) defined ICT as “those technologies that are used for accessing, gathering,

manipulating and presenting or communicating information” (p. 3). These technologies, she

explained, include hardware such as computers and other devices; software applications; and

connectivity such as access to the Internet. She then defined the use or integration of ICT in

pedagogy (UIP) as “a change in pedagogical approach to make ICT less peripheral to schooling

and more central to student learning” (p. 5). From its conception, UIP is of utmost importance.

According to Noor-Ul-Amin (2013), UIP provides opportunities to the learner to access an

abundance of information using multiple information resources and also to view information from

multiple perspectives.

Given its importance, UIP has attracted several studies. But how is UIP measured in those studies?

To that effect, several scholars (e.g. Jamieson-Proctor, Watson, Finger, Grimbeek & Burnett, 2006;

Papanastasiou & Angeli, 2008; Peeraer & Petegem, 2012; Proctor, Watson & Finger, 2003;

Puentedura, 2006) have made attempts to develop and/ or test instruments that measure UIP by



13

teachers and students. Of particular interest in this study is the substitution, augmentation,

modification and redefinition (SAMR) model (Puentedura, 2006). This study evaluated the validity

and reliability of an instrument based on the SAMR model in the context of the teachers of

mathematical disciplines in four universities in Uganda.

2. LITERATURE REVIEW

2.1 Instruments on the use of ICT in pedagogy than those based on SAMR

Jamieson-Proctor et al. (2006) developed an instrument that they termed the “Learning with ICTs:

Measuring ICT Use in the Curriculum”. Their instrument was an improvement of the “ICT

Curriculum Integration Performance Measurement” instrument (Proctor et al., 2003), which had

137 self-rated items, scaled using the four-point Likert (as Never, Sometimes, Often & Very often).

Using face validity that involved examining the items of the instrument for redundancy and

ambiguity, Jamieson-Proctor et al. reduced the items to 45 in number. They then administered the

refined instrument to 929 teachers in 38 state primary and secondary schools in Queensland,

Australia. Using a series of validity and reliability analyses, they further reduced the instrument to

20 items, and termed it “Learning with ICTs: Measuring ICT use in the Curriculum”.

Papanastasiou and Angeli (2008) constructed what they termed a “survey of factors affecting

teachers teaching with technology (SFA-T3)” (p. 70). Such factors included those related to (a) a

teacher’s knowledge of technology tools (14 items), and (b) a teacher’s frequency of using

technology for personal purposes (15 items). While they referred to their tool as being on “factors

affecting teachers teaching with technology”, in reality the tool had constructs – see those given

above – dealing with teaching with technology itself. They sought to determine the reliability and

validity of the instrument. Hence using Cronbach and factor analyses respectively, they found their

instrument sound.


Peeraer and Petegem (2012) developed an instrument to measure the integration of ICT in

education. Having carried out a literature search on the definitions of the integration of ICT in

education, they used the item response modeling approach (Wilson, 2005 cited in Peeraer &

Petegem, 2012, p. 1252) to develop a self-report instrument on teacher educators’ use of ICT for

teaching and support of students’ learning. In the bid to test the validity and reliability of their 15-

item instrument, Peeraer and Petegem (2012) collected data from 933 teacher educators working

in five teacher education institutions in five different regions in the north and centre of Vietnam.

Finally, using the Rasch model of measurement (Linacre, 2002 cited in Peeraer & Petegem, 2012),

they concluded that their instrument could be used with confidence.

2.2 Instruments on the use of ICT in pedagogy based on SAMR

As part of his work with the Maine Learning Technologies Initiative, Puentedura (2006) developed

the substitution, augmentation, modification and redefinition (SAMR) model of ICT which he

intended to encourage educators to significantly enhance the quality of education provided via

technology (Romrell, Kidder & Wood, 2014). According to Puentedura (2006 cited in Romrell et

al., 2014), at the level of substitution, ICT acts as a substitute for another technology, but with no

functional change. An example of this is when one uses a computer as a substitute for a type writer

for purposes of producing documents but without any significant change to the function of a type

writer. At the level of augmentation, ICT acts as a substitute for another technology, but with

functional improvement. For example, one can use a computer as a substitute for a type writer but

with significant functionality increase such as spell checking. At the modification level, ICT allows

for activities to be redesigned. An example in the modification dimension is when ICT allows for

the learning processes to integrate them with technology such as email and graphics packages.

Lastly, at the redefinition level, ICT allows for the creation of new tasks that were previously

inconceivable. According to Lubega et al. (2014), such tasks involve the use of visualization and

simulation tools as part of the learning activities. According to Puentedura (2013, cited in Romrell,

2014, p. 5), “learning activities that fall within the substitution and augmentation [S & A]

classification are said to enhance learning, while learning activities that fall within the modification

and redefinition [M & R] classifications are said to transform learning”.


15

In accordance with Green (2014) who contends that SAMR “is an extremely popular model” (p.

37), several researchers have used it to guide their studies on the use of ICT in pedagogy. For

example, Azama (2015) examined the integration of ICT by a beginning Japanese class in a public

high school in a rural town in California. He measured the integration of ICT using the SAMR

model. In particular, he reported that “technology enhanced activities… [were] developed based

on the SAMR model and students’ learning… documented through formative assessments and

personal reflections” (p. 21). In terms of results, Azama (2015) reported that,

although students’ responses did not show any particular SAMR stage being more engaging than

other stages, when students were asked to rate which technology tools they imagined they would

use in the future, significantly more students chose redefinition tasks over the other stages of

activities. Although students found some activities in the augmentation stage enjoyable, they

perceived those activities… as [mere] “tools of learning”. However, the activities in the

modification and redefinition stage[s] were perceived as useful by students for everyday purposes

(pp. 34 - 35).

Kihoza, Zlotnikova, Bada and Kalegele (2016) examined the technological knowledge,

competencies, skills, attitude, beliefs and readiness to integrate classroom technology of teacher

trainees and tutors. The trainees and tutors were from Morogoro Teacher Training College and

Mzumbe University both of Morogoro Region in Tanzania. They used an instrument structured

according to SAMR to collect data on the use of ICT in pedagogy from 206 respondents and used

descriptive statistics to show that the majority of the respondents had low pedagogical ICT

competencies. In particular they reported finding that, the “ICT competence level [that the]

majority reported was [that of] beginners” (p. 116). That is, the majority of the respondents were

at the substitution (S) level in SAMR terms. They went on to report that, in terms of the

“augmentation [A] attributes [that] were used to assess tutors’ and teacher trainees’ readiness to

integrate ICTs in classrooms”, the majority of the tutors reported being well prepared and

somewhat prepared. However for the teacher trainees, the majority reported either being poorly

prepared or not prepared at all (p. 117). They recorded similar patterns for the use of ICT at the

modification (M) and redefinition (R) levels.


Lubega et al. (2014) assessed the adoption of ICT in pedagogy by the academic staff and students

of Makerere University in Uganda. Using an instrument that they had structured as per the SAMR

model, they collected data from 600 respondents, which they analysed using descriptive statistics,

notably percentages. Hence Lubega et al. found that non-use of a number of ICT tools in

pedagogical processes in Makerere was highly prevalent. In particular, at the substitution (S) level

in SAMR terms, they reported finding that, “a number of teaching activities... [were] yet to be

computerized even at the basic substitution level” (p. 110). Then at the augmentation (A) level,

they summarized their finding by noting that, “more pedagogical activities in and outside the

classroom... [were] mainly ported onto substitution ICTs than augmentation ICTs” (p. 111). At the

modification (M) stage, they reported that, “the most common modification ICT... [was] the

Internet” (p. 112). Lastly with regard to the use of ICT at redefinition (R) level, they reported only

a trace of activities at that level.

Speirs (2016) assessed mobile device skills and the integration of technology into the lives of pre-

service teachers at the State University New York (SUNY) Plattsburgh and practicing teachers in

New York State. Having structured his questionnaire using SAMR to categorise activities

commonly completed with mobile devices, Speirs (2016) used a class of 10 psychometrics

graduate students and their professor to ensure the face validity of the items. Then he collected

data from 53 respondents. In terms of analysis, he reported that “total values for each SAMR level

were calculated in accordance with the structure of the survey” (p. 15). Also, “scores for total

integration level (IntTot)… were evaluated for normality” (p. 16). Hence in terms of findings, he

reported that the “skewness for IntTot scores was highly… to the left, indicating that many

participants indicated high level[s] of technology integration into their lives” (p. 18).

Apart from empirical studies using SAMR, even literature reviews (e.g. Green, 2014; Romrell et

al., 2014) on the same are available. Green (2014) started generously by noting that “SAMR is an

extremely popular model” (p. 37) because “SAMR is clean and simple, which means it can be

easily adapted and interpreted in multiple ways” (p. 38). However she went on to question whether

Ruben Puentedura who had a PhD in Chemistry and whose name features “on several institutional

documents and news articles concerning his work in Physics, Chemistry and multimedia labs” (p.


17

38) could have originated the SAMR model independently. Green further asserted that,

“Puentedura has not published any results of the decade of study he claims to have conducted” (p.

39) as he developed SAMR. She thus went on to insinuate that Puentedura may have plagiarized

the idea of SAMR from Hughes (2005) whose article had identified three functions of technology

as replacement (which SAMR calls substitution), amplification (which SAMR calls augmentation)

and transformation (which terms SAMR reserves for both its modification and redefinition

constructs together). Another shortcoming of SAMR, according to Green (2014) was that Ruben

Puentedura who proposed it, had hardly used it in his empirical studies. In particular, Green (2014)

contended that, “no peer reviewed papers on this model have been authored and published by

Puentedura” (p. 38-39). That implied that it is others (e.g. Azama, 2015; Kihoza et al., 2016;

Lubega et al., 2014; Speirs, 2016) who have empirically used SAMR and not its proponent,

Puentedura, himself. Green (2014) however, made a general critique of all such studies lamenting

that, “there is no record of research studies that could document the development and validation

of SAMR” (p. 39). In other words, the studies that have used the SAMR model have used it at face

value without checking its validity and reliability.

Romrell et al. (2014) devoted their energies to highlighting the SAMR model as a framework for

evaluating mobile learning (mLearning). Having defined mLearning as “learning that is

personalized, situated, and connected through the use of a mobile device” (p. 1), they went on to

review recent literature on mLearning and to provide examples of activities from studies, that fell

within each of the four classifications of the SAMR model. The classifications, as expected, were

substitution, augmentation, modification and redefinition.

As suggested by the reviewed studies, as Green (2014) claimed, among the studies (e.g. Azama,

2015; Kihoza et al., 2016; Lubega et al., 2014; Speirs, 2016) that have used the SAMR model

empirically, “there is no record of research studies that could document the development and

validation of SAMR” (p. 39). Indeed, as we prepared this article, our efforts to come across one

scholarly article of that kind was in vain. In other words, the studies that have used the SAMR

model have used it at face value without checking its validity and reliability. On the basis of such

gaps, this study came in handy to test the validity and reliability of the SAMR model using the


teachers of mathematical disciplines in four universities in Uganda. In particular, the study sought;

first, to establish the validity and reliability of each of the four constructs (S, A, M & R). The

second objective of this study was to test whether the four constructs were independent. Third, the

study was to re-examine whether the SAMR model of ICT as being made up of the four constructs

(S, A, M & R) was reasonable.

3. METHODOLOGY

3.1 Instrument

Using the survey design, data were collected using Puentedura’s (2006) SAMR measure of the use

of ICT in pedagogy (UIP), which operationalises the use of ICT in pedagogy as having the

substitution (S), augmentation (A), modification (M), and redefinition (R) constructs. This

instrument was developed by Lubega et al. and hence sourced from Lubega et al. (2014, Tables

III, IV, V & VIII). However, before collecting the data, preliminary validation of the instrument

was carried out using face validity to see which items were applicable to teachers at university

level. Hence, the items on S reduced from 13 to 12; those on A reduced from 16 to nine; those on

M reduced from 10 to five; and lastly those on R reduced from six to five in number. In total there

were 31 items measuring UIP. The items were scaled using the five-point Likert scale from a

minimum of 1 for the worst case scenario (strongly disagree) to a maximum of 5, which was the

best case scenario (Strongly agree).

3.2 Sample

The sample comprised 261 academic staff teaching mathematical disciplines in four universities

in Uganda, namely Kyambogo (KyU), Makerere (Mak), Makerere University Business School

(MUBS) and Mountains of the Moon (MMU). The term “mathematical disciplines” was taken to

be broad and thus included a range of disciplines where mathematical skills are useful. Such

disciplines included Science, Technology, Engineering and Mathematics (STEM) disciplines and

their offshoots like Biomathematics, Finance, Management Science, Quantitative Data

Management, and Software Engineering. As suggested in Table 1, a typical respondent was aged

30 but below 40 years (62.9%); a male (64.0%); belonging to the Makerere University Business


19

School, MUBS (36.0%); having served for below five years as a lecturer at university level

(56.6%); holding a Masters degree (69.6%) as the highest qualification; and currently holding the

academic rank of Lecturer (48.8%).

Table 1: Background Characteristics of the Respondents

Item Categories Frequency Percent

Age group in years of the respondent Up to 30 33 15.7

30 but below 40 132 62.9

40 and above 45 21.4

Gender of the respondent Female 94 36.0

Male 167 64.0

University to which the respondent

belonged

Kyambogo

Makerere

50

87

19.2

33.3

Makerere University Business School 94 36.0

Mountains of the Moon 30 11.5

Tenure in years of lecturing at university

level of the respondent

Up to five 141 56.6

Five but below 10 85 34.1

10 and above 23 09.2

Highest academic qualification attained

by the respondent

Bachelors degree 20 07.7

Masters degree 181 69.6

PhD degree 59 22.7

Academic rank of the respondent Teaching Assistant

Assistant Lecturer

20

68

07.7

26.2

Lecturer 127 48.8

Senior Lecturer 37 14.2

Associate Professor 06 02.3

Professor 02 0.8

3.3 Data Analysis

The validities of multi-item constructs of SAMR, namely S, A, M and R, were tested using

confirmatory factor analysis (CFA), while their reliabilities were tested using the Cronbach alpha

method. Pearson's linear correlation analysis was carried out to establish whether the constructs of

SAMR were independent, while exploratory factor analysis (EFA) helped with the re-assessment

of the structure of SAMR.

4. RESULTS

4.1 Validities and Reliabilities


The first objective of the study was to establish the validity and reliability of the measure for each

of the four constructs of SAMR. This was achieved via confirmatory factor analysis (CFA) and

Cronbach alpha on the items on each construct. The Kaiser rule or criterion (Kaiser, 1960 cited in

Khan, 2006, p. 690; Yong & Pearce, 2013, p. 85) or the Kaiser-Guttman rule (cited in Schmidt,

Baran, Thompson, Mishra, Koehler & Shin, 2009, p. 131) that stipulates that factors with

eigenvalues greater than one be considered significant, was used in the study. For a given factor

loading, 0.5 (Costello & Osborne, 2005) was used as the minimum. For reliability tests, a

benchmark of α = 0.7 (Tavakol & Dennick, 2011) for the Cronbach alpha was set. The results are

as presented in the subsequent tables (Tables 2 through 5).

4.1.1 Use of Substitution ICTs

According to Table 2, confirmatory factor analysis (CFA) reduced the 12 items of the first

construct (S) in SAMR to three factors. The respective factors had eigenvalues of 6.42, 1.54 and

1.16, meaning that the respective three factors accounted for 6.42/ 12 x 100 = 53.5%; 1.54/ 12 x

100 = 12.8%; and 1.16/ 12 x 100 = 09.7% of the total variance among the 12 items. The loadings

of the respective items on a given factor are also given in Table 2 after a Varimax rotation, as

recommended by Kline (1994, cited in Foster, 1998, p. 207). Considering loadings of at least 0.5

as being high, then Table 2 suggests that only seven items (S3 – S5, S7 – S10) loaded highly on

the first and main factor. Five items (S1, S2, S4 – S6) loaded highly on the second factor; while

two others (S11 & S12) loaded highly on the third and least among the significant factors. Given

that “Kaiser’s criterion will often yield too many factors [to be] retained” (Khan, 2006, p. 690),

for the sake of parsimony, as advised by Khan (2006, pp. 690 - 691), only the five items (S3, S7 –

S10) that loaded highly on only the first and main factor, were taken as the (most) valid items of

S. Items S4 and S5 cross-loaded on the first two factors and hence were dropped for being complex

(Yong & Pearce, 2013, p. 84). The reliability (Cronbach alpha, α) of the valid items (S3, S7 – S10)

was 0.920, which being greater than 0.7, implied that the five items were reliable measures of S

too.


21

Table 2: Loadings on Factors on the Use of Substitution ICTs (S)

Loadings

Items Descriptions Factor 1 Factor 2 Factor 3 α

S1 I use ICTs to prepare my lecture notes, assignments and

examinations

0.910 0.920

S2 I use presentation software (e.g. PowerPoint) to deliver my

lectures

0.709

S3* I upload my teaching and learning materials on electronic

sites/ devices (e.g. MUELE) for students to access

0.711

S4 When supporting my students, I communicate to them

using e-mail

0.571 0.662

S5 I refer my students to electronic databases for reference

materials

0.569 0.634

S6 When supporting my students, I communicate to them

using my cell phone

0.570

S7* During my lectures, I use the smart boards/ interactive

boards installed in the lecture rooms for writing

0.860

S8* I encourage students to submit their course work

assignments through e-mail

0.800

S9* When communicating to my students, I use electronic

notice boards

0.899

S10* When supporting my students, I communicate to them

through social media (e.g. blogs, chat rooms, discussion

boards, Facebook, Instagram and Twitter)

0.796

S11 I record my lectures on CDs or other media and give them

to my students

0.850

S12 I take video/ audio recordings of myself while lecturing

and use them in subsequent years to teach the same course

to another cohort of students

0.908

Eigenvalue 6.417 1.536 1.160

% variation

explained

53.5 12.8 09.7

*Valid items of S

4.1.2 Use of Augmentation ICTs

According to Table 3, CFA reduced the nine items of the second construct (A) in SAMR to two

factors. The respective factors had eigenvalues of 5.07 and 1.17, meaning that the respective two


factors accounted for 56.3% and 13.0% of the total variance among the nine items. The loadings

of the respective items on a given factor are also given in Table 3 after a Varimax rotation.

Considering loadings of at least 0.5 as being high, then Table 3 suggests that only five items (A5

– A9) loaded highly on the first and main factor. The other four items (A1 - A4) loaded highly on

the second and minor among the significant factors. As was the case for the construct S, and for

the same reasons, only the five items (A5 – A9) that loaded highly on the first and main factor,

were taken as the (most) valid items of A. Their reliability (Cronbach alpha, α) of 0.861 implied

that the five items were also reliable measures of A.

Table 3: Loadings on Factors on the Use of Augmentation ICTs (A)

Loadings

Items Descriptions Factor 1 Factor 2 α

A1 I use search engines (e.g. Google) to look for content in my

discipline

0.834 0.861

A2 I use editorial tools (e.g. spell checker) in my word processor

to correct grammatical errors in any document I process

0.847

A3

I use editorial tools (e.g. Thesaurus) in my word processor to

get alternative words to use in my documents

0.717

A4

I use online encyclopedias (e.g. Wikipedia) to make meaning

of words or phrases that I do not understand

0.695

A5*

I use digital libraries (e.g. MuLib and MakIR) as a source of

useful content for my lectures

0.781

A6*

I use the track changes tool in my word processor to review

documents (e.g. student dissertations/ theses)

0.645

A7*

I use internet group lists to contact my students in matters

related to their academics

0.785

A8*

I encourage my students to use Google docs to accomplish

group course work assignments

0.795

0.811

A9*

I use different videos to illustrate different case studies during

my lectures

0.745

Eigenvalue

5.069

1.167

% variation

explained

56.3

13.0

*Valid items of A

4.1.3 Use of Modification ICTs


23

According to Table 4, CFA reduced the five items of the third construct (M) in SAMR to the ideal

situation of one factor. The factor had an eigenvalue of 3.08, meaning that the factor accounted for

61.5% of the total variance among the five items. The loadings of the respective items on the factor

are also given in Table 4. Considering loadings of at least 0.5 as being high, then Table 4 suggests

that all the five items (M1 – M5) loaded highly on the factor. Hence all of them were valid items

of M. Their reliability (Cronbach alpha, α) was 0.842, implying that the five items were also

reliable measures of M.

Table 4: Loadings on the Factor on the Use of Modification ICTs (M)

Items * Descriptions Loadings α

M1 I assign students topics to research about from the Internet 0.517 0.842

M2 I lecture modules in my discipline using e-learning platforms

(e.g. MUELE)

0.892

M3 I use content authoring software when preparing my lectures 0.882

M4 I use online tools (e.g. RM Assessor) to assess my students 0.898

M5 I use video conferencing or Skype to teach my students when I

am not at the University

0.655

Eigenvalue 3.075

% variation

explained

61.5

* All items were valid for M

4.1.4 Use of Redefinition ICTs

According to Table 5, CFA reduced the five items of the fourth and last construct (R) in SAMR to

the ideal situation of one factor. The factor had an eigenvalue of 3.46, meaning that the factor

accounted for 69.2% of the total variance among the five items. The loadings of the respective

items on the factor are also given in Table 5. Considering loadings of at least 0.5 as being high,

then Table 5 suggests that all the five items (R1 – R5) loaded highly on the factor. Hence all of

them were valid items of R. Their reliability (Cronbach alpha, α) was 0.888, implying that the five

items were reliable measures of R too.

Table 5: Loadings on the Factor on the Use of Redefinition ICTs (R)

Items * Descriptions Loadings α

R1 I use open education resources (e.g. massive open

online courses, MOOCS) as my lecturing materials

0.851 0.888

R2 I use electronic games when lecturing 0.514

R3 I use simulations (e.g. 2nd life) when lecturing 0.903


R4

I use e-learning platforms (e.g. MUELE) to encourage

group discussions among my students

0.900

R5

I use e-learning platforms (e.g.MUELE) to assess my

students’ learning

0.921

Eigenvalue 3.460

% variation

explained

69.2

*All items were valid for R

4.2 Correlations among the SAMR Constructs

The second objective of the study was to test whether the four constructs in the SAMR model,

namely substitution (S), augmentation (A), modification (M) and redefinition ( R) were

independent. Average indexes were computed for the valid items of the respective constructs from

Tables 2 - 5. The indexes were then correlated using Pearson's linear correlation. Table 6

(correlation matrix) suggests that all the four constructs were significantly inter-related.

Table 6: Inter-correlations of the SAMR Constructs

S A M R

S

A 0.667**

M 0.783** 0.727**

R 0.770** 0.703** 0.864**

** Correlation is significant at the 0.01 level

4.3 Re-examining the Structure of SAMR

The third and last objective in the study was to re-examine whether the SAMR model of ICT

(Puentedura, 2006) as being made up of the four constructs (S, A, M & R) was reasonable.

Exploratory factor analysis (EFA) reduced the 31 items in the SAMR instrument (Tables 2 - 5)

into as many factors. However, as suggested in Table 7, only the first six factors were significant

since they had eigenvalues = 15.17, 2.40, 2.32, 1.25, 1.12 and 1.08 that exceeded 1.00. These

factors explained 48.9%, 7.7%, 7.5%, 4.0%, 3.6% and 3.5% respectively of the joint variation in

the 31 items. The items with high factor loadings (of at least 0.5) of the respective factors are also

given in Table 7 after a Varimax rotation. The question is: Is the SAMR structure as suggested by


25

Puentedura (2006) discernible in Table 7? Table 7 suggests that only the first four of the factors

were significant given that the sixth one had only two cross-loading and hence complex items. The

fifth one had only two valid items and could be disregarded, in accordance with Yong and Pearce

(2013) who observe that, “factors that have less than three variables… are… undesirable” (p. 86).

Thus in agreement with the SAMR model, Table 7 suggested four factors from the 31 items on the

use of ICT.

Apart from Factor 3 that was dominated by only one construct (A), none of the other three

significant factors was dominated by items from only one construct of SAMR. In particular, Factor

1 had five valid items (S3, S7, S8, S9 & S10) on the use of substitution ICTs; one valid item (A7)

on the use of augmentation ICTs; three valid items (M2 - M4) on the use of modification ICTs;

and three valid items (R3 – R5) on the use of redefinition ICTs. Factor 2 had two valid items (S5

& S6) on the use of substitution ICTs; and three valid items (A5, A6 & A8) on the use of

augmentation ICTs. Factor 4 had two valid items (S11 & S12) on the use of substitution ICTs; and

one valid item (A9) on the use of augmentation ICTs. In summary, the SAMR structure as

suggested by Puentedura (2006) could hardly be discerned in Table 7.

Table 7 Factors, their eigenvalues, % variance explained and their highly loading

items

Factor Eigenvalue % variance Highly loading items (loading in brackets)

1 15.17 48.9 S3 (0.689); *S4 (0.504); S7 (0.792); S8 (0.717); S9 (0.847); S10 (0.696);

** A4 (0.500); A7 (0.754); M2 (0.868); M3 (0.729); M4 (0.697); ***R1

(0.540); R3 (0.691); R4 (0.868); R5 (0.885)

2 2.40 7.7 *S4 (0.626); S5 (0.643); S6 (0.647); A5 (0.781); A6 (0.628); A8 (0.624)

3 2.32 7.5 A1 (0.751); A2 (0.762); A3 (0.692); ** A4 (0.609)

4 1.25 4.0 S11 (0.799); S12 (0.893); A9 (0.531); ****R2 (0.529)

5 1.12 3.6 S1 (0.762); S2 (0.771)

6 1.08 3.5 ***R1 (0.600); ****R2 (0.583)

Footnote: Items prefixed with a similar symbol (*S4, **A4, ***R1 and ****R2 ) were cross-loading, and hence dropped for being

complex

5. DISCUSSION

The first objective in the study was to establish the validity and reliability of each of the four

constructs, namely substitution (S), augmentation (A), modification (M) and redefinition ( R) in


Puentedura's (2006) SAMR model of ICT. Confirmatory factor analysis (CFA) showed that the

items on construct S were categorisable into three factors (see Table 2, i.e. S3, S7 – S10 as the

most important factor; S1 and S6 as another less significant factor; and S11 and S12 forming the

third and least significant factor). This means that if for example all the items on S were to be an

independent variable in a model, it would automatically mean that the dependent variable would

be regressed on three different factors on S and not one as suggested by the SAMR model. If it

was a question of looking for the valid items on construct S, then as we did in this paper (subsection

4.1.1), those factors (S3, S7 – S10) that loaded highly on only the first and main factor, were the

most optimal set, given their high reliability (Cronbach alpha, α) of 0.898. Dropping the other

seven items (S1, S2, S4 – S6, S11 & S12) in favour of only five is implying that the construct S of

SAMR may be unnecessarily long.

CFA showed that the items on construct A were categorisable into two factors (see Table 3, i.e.

A5 – A9 as the most important factor; and A1 – A4 forming a second but less important factor).

This means that if for example all the items on A were to be an independent variable in a model,

it would automatically mean that the dependent variable would be regressed on two different

factors on A and not one as suggested by the SAMR model. If it was a question of looking for the

valid items on construct A, then as we did in this paper (subsection 4.1.2), those factors (A5 – A9)

that loaded highly on only the first and main factor, were the most optimal set, given their high

reliability (Cronbach alpha, α) of 0.861. Dropping the other four items (A1 – A4) in favour of only

five is implying that the construct A of SAMR may be unnecessarily long. CFA showed that the

items on construct M could be reduced to one factor (see Table 4). This meant that all the five

items on construct M that loaded highly on the one factor, were valid items of M. They were also

reliable, given their high reliability (Cronbach alpha, α) of 0.842. That all the five items (M1 –

M5) were valid and reliable, might be implying that the construct M of SAMR may be optimally

long. CFA showed that the items on construct R could also be reduced to one factor (see Table 5).

Also all the five items (R1 – R5) loaded highly on the factor, implying that they were valid ones.

They were also reliable (their Cronbach alpha, α = 0.888), suggesting that they were the ideal

measure of R.


27

The second objective in the study was to test whether the four constructs of SAMR, namely S, A,

M and R were independent. The results of correlation analysis (Table 6) suggested that all the four

constructs were inter-related. This puts into question whether they are really measuring different

things. That is food for thought for future researchers. It could also imply that when carrying out

a study using the constructs of SAMR as explanatory variables, the researcher does not have to

include all of them. The third and last objective of the study was to re-examine whether the four-

factor SAMR model of ICT, that is, as made up of the four constructs (S, A, M & R) was

reasonable. Exploratory factor analysis, EFA (Table 7) showed that the SAMR structure as

suggested by Puentedura (2006) and operationalised by Lubega et al. (2014), could hardly be

replicated using our data set. While our exploratory study using teachers of mathematical

disciplines in four universities in Uganda cannot be used to dismiss Puentedura’s (2006)

suggestions and the operationalisation of Lubega et al. (2014), it still raises a lot of food for thought

for future researchers. Is the four-factor SAMR model really reasonable? This question still begs

for research in several contexts, and from researchers including Ruben Puentedura himself that

Green (2014) accused of hardly having answered the question.

6. CONCLUSION

The use of ICT in pedagogy (UIP) provides opportunities to the learner to access an abundance of

information using multiple information resources and also to view information from multiple

perspectives. As such, UIP increases learner motivation and engagement by facilitating the

acquisition of basic skills (Noor-Ul-Amin, 2013). Given the importance of UIP, in the study being

concluded, an attempt was made to test the validity and reliability of one of the instruments that

measure UIP by teachers and students. The instrument was the substitution (A), augmentation (A),

modification (M) and redefinition (R) - SAMR model (Puentedura, 2006, as operationalized by

Lubega et al., 2014). The context was that of the teachers of mathematical disciplines in four

universities in Uganda. The study was among the very first to report findings on the validity and

reliability of the SAMR model of ICT. CFA suggested that while some items of S and A

respectively were not valid, all the items of each of M and R were valid. The four constructs were

highly inter-related. Exploratory factor analysis (EFA) revealed that the four-factor SAMR model


of operationalising ICT was questionable. Hence a call was made to researchers to continue testing

the SAMR model in different contexts with intent to refining it. The study had limitations.

Obviously, the sample scope was limited. More studies on SAMR could be carried out among

other teachers in the four universities than those of mathematical disciplines. Studies could even

be extended to other universities in Uganda and other countries.

REFERENCES

AZAMA, Y. 2015. Effective integration of technology in a high school beginning Japanese

class. Capstones and Theses at Digital Commons. Paper 517. Retrieved from:

http://digitalcommons.csumb.edu/caps-theses.

COSTELLO, A. B., AND OSBORNE, J. W. 2005. Best practices in exploratory factor analysis:

Four recommendations for getting the most from your analysis. Practical Assessment Research

and Evaluation, 10 (7). Retrieved from: http://pareonline.net/getvn.asp?v=10&n=7.

FOSTER, J. J. 1998. Data analysis using SPSS for Windows. London, UK: Sage.

GREEN, L. S. 2014. Through the looking glass: Examining technology integration in school

librarianship. Knowledge Quest, 43(1), 36-43.

HUGHES, J. 2005. The role of teacher knowledge and learning experiences in forming

technology-related pedagogy. Journal of Technology and Teacher Education, 13(2), 277-302.

JAMIESON-PROCTOR, R. M., WATSON, G., FINGER, G., GRIMBEEK, P., & BURNETT,

P. C. 2006. Measuring the use of information and communication technologies (ICTs) in the

classroom. Computers in the Schools, 24(1), 167-184.

http://digitalcommons.csumb.edu/caps-theses

http://pareonline.net/getvn.asp?v=10&n=7


29

KAISER, H. F. 1960. The application of electronic computers to factor analysis. Educational

and Psychological Measurement, 20, 141 - 151. doi: 10.1177/001316446002000116

KHAN, J. H. 2006. Factor analysis in counseling psychology research, training, and practice:

Principles, advances, and applications. The Counseling Psychologist, 34 (5), 684 - 718. doi:

10.1177/0011000006286347

KIHOZA, P., ZLOTNIKOVA, I., BADA, J., AND KALEGELE, K. 2016. Classroom ICT

integration in Tanzania: Opportunities and challenges from the perspectives of TPACK and SAMR

models. International Journal of Education and Development Using Information and

Communication Technology, 12(1), 107-128.

KLINE, P. 1994. An easy guide to factor analysis. New York, US: Routledge.

LINACRE, J. M. 2002. Understanding Rasch measurement: Optimising rating scale category

effectiveness. Journal of Applied Measurement, 31 (1), 85 - 106.

LLOYD, M. 2005. Towards a definition of the integration of ICT in the classroom. In AARE

(Eds.), Proceedings of AARE ‘05 on Education research – creative dissent: Constructive solutions.

Parramatta, New South Wales.

LUBEGA, T. J., MUGISHA, A. K., AND MUYINDA, P. B. 2014. Adoption of the SAMR

model to assess ICT pedagogical adoption: A case of Makerere University. International Journal

of e-Education, e-Business, e-Management and e-Learning, 4(2), 106-115.

NOOR-UL-AMIN, S. 2013. An effective use of ICT for education and learning by drawing on

worldwide knowledge, research and experience: ICT as a change agent for education. Scholarly

Journal of Education, 2(4), 38-54.

PAPANASTASIOU, E. C., AND ANGELI, C. 2008. Evaluating the use of ICT in education:


Psychometric properties of the survey of factors affecting teachers teaching with technology (SFA-

T3). Journal of Educational Technology & Society, 11 (1), 69 - 86. Retrieved from:

http://www.jstor.org/stable/jeductechsoci.11.1.69.

PEERAER, J., AND PETEGEM, P. V. 2012. Measuring integration of information and

communication technology in education: An item response modeling approach. Computers &

Education, 58, 1247-1259.

PROCTOR, R., WATSON, G., AND FINGER, G. 2003. Measuring information and

communication technology (ICT) curriculum integration. Computers in Schools, 20(4), 67-87.

PUENTEDURA, R. R. (2006, November 28). Transformation, technology, and education in the

state of Maine. [Web log post]. Retrieved from:

http://www.hippasus.com/rrpweblog/archives/000095.html.

PUENTEDURA, R. R. 2013, May 29. SAMR: Moving from enhancement to transformation.

[Web log post]. Retrieved from: http://www.hippasus.com/rrpweblog/archives/000095.html.

ROMRELL, D., KIDDER, L. C., AND WOOD, E. 2014. The SAMR model as a framework

for evaluating mlearning. Journal of Asynchronous Learning Networks, 18(2), 1-15.

SCHMIDT, D. A., BARAN, E., THOMPSON, A. D., MISHRA, P., KOEHLER, M. J., AND

SHIN, T. S. 2009. Technological pedagogical content knowledge (TPACK): The development and

validation of an assessment instrument for preservice teachers. Journal of Research on Technology

in Education, 42(2), 123-149.

SPEIRS, M. N. 2016. Assessing teachers’ mobile device skills and the integration of

technology into their lives. Psychology at Digital Commons. Paper 6. Retrieved from:

http://digitalcommons.plattsburgh.edu/psychology-theses.

TAVAKOL, M., AND DENNICK, R. 2011. Making sense of Cronbach’s alpha. International

http://www.jstor.org/stable/jeductechsoci.11.1.69

http://www.hippasus.com/rrpweblog/archives/000095.html

http://www.hippasus.com/rrpweblog/archives/000095.html

http://digitalcommons.plattsburgh.edu/psychology-theses


31

Journal of Medical Education, 2, 53-55. doi: 10.5116/ijme.4dfb.8dfd

WILSON, M. 2005. Constructing measures: An item response modeling approach. Mahway:

Erlbaum.

YONG, A. G., AND PEARCE, S. 2013. A beginner’s guide to factor analysis: Focusing on

exploratory factor analysis. Tutorials in Quantitative Methods for Psychology, 9(2), 79-94.

Empirical Study of Continuous Change of Open Source System

Michael Abayomi Olatunji3, Rufus Olalere Oladele and Amos Orenyi Bajeh

Department of Computer Science

University of Ilorin, Ilorin, Nigeria

P. M. B. 1515, Ilorin, Nigeria.

[email protected], [email protected], [email protected]

___________________________________________________________________________

ABSTRACT

This paper describes empirical investigation of the first law of software evolution: law of

continuing change, using four open source systems in evolution as case studies. The four systems

used as case studies have 354 versions altogether with a combined 55 years of evolution and

running existence. Three code analysing tools were used to parse source codes and generated

3 Author’s Address: Michael Abayomi Olatunji3, Rufus Olalere Oladele and Amos Orenyi Bajeh

Department of Computer Science, University of Ilorin, Ilorin, Nigeria, P. M. B. 1515, Ilorin, Nigeria.

[email protected], [email protected], [email protected]"Permission to make digital or hard copies of part or all of this work for

personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that

copies bear this notice and the full citation on the first page. Copyrights for components of this work owned by others than IJCIR must be

honored. Abstracting with credit is permitted. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific

permission and/or a fee." © International Journal of Computing and ICT Research 2008. International Journal of Computing and ICT Research,

ISSN 1818-1139 (Print), ISSN 1996-1065 (Online), Vol.11, Issue 2, pp. 31 - 52, June 2017.


metrics needed for quantification of continuing change property. Statistical functions and

descriptive statistics were used for metrics processing and analysis. Results of the study confirm

the law of continuing change. In particular, results show that: (i) in the pre versions, lot of

changes do occur but when it is getting close to first release candidate of version 1, the changes

reduce but in stable branch after the first official release more changes do occur close to the

release of another stable or development major release. (ii) More changes occur in the function

body than in the function interfaces. (iii) Changes in the stable branches are fewer than those in

the development branches. (iv) Function interface changes are correlated to the function

statement changes. Within the domain of the case studies investigated, there is no system that

does not have shrinkages in number of functions and Lines of code in the course of evolution.

Keywords: Lehman Law, Empirical investigation, Open Source Systems, Software Metrics,

Software Evolution, Quantification.


Michael Abayomi Olatunji4, Rufus Olalere Oladele and Amos Orenyi Bajeh. Empirical Study of

Continuous Change of Open Source System, Vol. 11, Issue.1 pp 31 - 52.


INTRODUCTION

4 Author’s Address: Michael Abayomi Olatunji4, Rufus Olalere Oladele and Amos Orenyi Bajeh

Department of Computer Science, University of Ilorin, Ilorin, Nigeria, P. M. B. 1515, Ilorin, Nigeria.

[email protected], [email protected], [email protected]"Permission to make digital or hard copies of part or all of this work for

personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that

copies bear this notice and the full citation on the first page. Copyrights for components of this work owned by others than IJCIR must be

honored. Abstracting with credit is permitted. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific

permission and/or a fee." © International Journal of Computing and ICT Research 2008. International Journal of Computing and ICT Research,

ISSN 1818-1139 (Print), ISSN 1996-1065 (Online), Vol.11, Issue 2, pp. 31 - 52, June 2017.



33

Software evolution is the process by which softwares are modified and adapted to their changing

environment to maintain relevance. Software evolution became a field of research owing to the

pioneering empirical studies done by Meir Lehman, which he started by using IBM systems as

case studies about 47 years ago. Lehman aimed at formulating theory for software evolution from

empirical and statistical perspective. He intended to bring out commonalities or invariants within

software systems. In the course of pursuing the above aim, he came by three debatable invariants

which were postulated as the laws of software evolution in 1974. After years of continual

research in software evolution, more invariants were observed and postulated as laws; by 1996,

the laws of software evolution became eight in number, amongst which is the law of Continuing

Change which is further investigated in this study. The laws of software evolution are said to be

referring only to software systems classified as Evolutionary type systems (E-type systems),

these are systems that solve problems involving people or real world. They are reflections of

human processes. E-type systems can be Open Source Systems (OSS) or commercial/proprietary

systems which are close source systems developed for a particular organisation.

The law of continuing change was the first of the three laws of software evolution postulated in

1974 and it states that E-type systems must be continually adapted else they become

progressively less satisfactory [Lehman 96b]. Since user requirements changes with time,

software systems are inevitably subject to change and thus, must evolve to continually meet the

needs of the users. Otherwise, with time the software will not be meeting users' needs and

satisfaction, thereby losing relevance. The other laws of software evolution which are not the

focus of this paper are stated as follows:

Increasing complexity (1974): as an E-type system evolves its complexity increases unless work

is done to maintain or reduce it.

Self-regulation (1974): E-type system evolution process is self-regulating with distribution of

product and process measures close to normal.

Conservation of organisational stability (1980): the average effective global activity rate in an

evolving E-type system is invariant over product lifetime (Invariant work rate).


Conservation of familiarity (1980): as an E-type system evolves all associated with it, developers,

sales personnel, users, for example must maintain mastery of its content and behaviour to achieve

satisfactory evolution. Excessive growth diminishes that mastery. Hence the average incremental

growth remains invariants as the system evolves.

Continuing growth (1980): the functional content of E-type system must be continually increased

to maintain satisfaction over its lifetime.

Declining Quality (1996): The quality of E-type systems will appear to be declining unless they

are rigorously maintained and adapted to the operational environment changes.

Feedback system (first stated in 1974, formalised in 1996): E-type system evolution process

constitute multi-level, multi-loop, multi agent feedback system and must be treated as such to

achieve significant improvement over reasonable base.

Lehman studied seven commercial large software systems and came out with the laws of software

evolution. There are lot of differences between commercial systems and OSS. For instance, the

development of OSS is done by programmers scattered over the world. Also, the development of

OSS and its evolution/maintenance are been done at same time which happen at different times in

commercial systems. Some studies [Godfrey and Tu, 2000; Herraiz 2008; Fernandez-Ramil et al.

2008; Israeli and Feitelson 2010; Neamtiu et al. 2013; Pirzada 1988; Alenezi and Almustafa 2015;

Eick et al. 2001; Vasa, 2010] show that some of the laws are not holding or partially hold in OSS.

The laws have to be investigated with OSS in order to check the applicability of the laws with OSS

and to arrive at theory for software evolution, most especially with the availability of lot of open

source codes in diverse repositories on internet. This paper is an attempt towards validating the

first law, law of continuing change, using a new set of metrics.

RELATED WORKS

Since the promulgation of the Lehman laws [Lehman 74; Lehman 78; Lehman 79; Lehman 80;

Lehman 85a; Lehman 85b; Lehman 85c; Lehman 89; Lehman 90; Lehman 96], several studies

have been carried out that attempts to validate and also refine the laws. The last expanded version


35

of the laws published in 1996 [Lehman 96b] remains unmodified since then. However, it was since

1996 that most studies questioning or confirming their validity have been published.

Lawrence [1982] in his study on software evolution dynamics, validated the law of continuing

change using number of modules handed in each release as metric.

Pirzada [1988] addressed the validity issue of the Lehman's laws of software evolution using

statistical approach in his PhD thesis. In the study, the evolution of several flavours of UNIX was

investigated. One of the results of Pirzada's study showed that all the UNIX flavours verified the

first law of continuing change. He considered number of modules in release as metric.

Godfrey and Tu [2000] studied the case of the Linux kernel in an attempt to verify the laws of

software evolution in the context of Open Source Software (OSS) and found out that it was

evolving at an accelerating pace and obeys the law of continuing change as well. Godfrey and Tu

used system and module size as metrics for quantifying the Law of Continuing Change.

Koch [2007] addressed the validity question with a large sample of software projects, precisely

8,621 projects from SourceForge.net repository. Koch used SLOC as metric and graphical

statistical evaluation. He observed that most of the projects grow either linearly or with a

decreasing rate, according to the laws. But, about 40% of the projects showed a super linear pattern

which disincline with the laws. Koch speculated that the cause of the super linear growth might be

a certain organizational model, that he called the chief programmer team, a term originated in IBM

in the seventies.

Vasa [2010] studied 40 large open source systems using size and complexity metrics at the class

level in his study. Inclinations were found for the law of continuing change, law of self-regulation,

law of conservation of familiarity and law of continuing Growth.

More so, Israeli and Feitelson [2010] also studied the Linux kernel, they investigated if they fulfil

laws or not. They found that linear kernel growth in evolution is initially super linear and later

became linear from 2.5 release. Therefore, the law of continuing change is verified but at a

different change rate. Israeli and Feitelson concluded that Linux confirmed most of the Lehman's


laws; especially those of growth, continuing change and stability. They used number of source

files counted in arch and drivers directories of Linux kernel as metric to quantify continuing change

law.

In a study of 705 releases of 9 open source software projects, Neamtiu et al. [2013] reported that

only the laws of continuing change and continuing growth are confirmed for all programs. Neamtiu

research group are the collectors of the datasets used in this study. In their study, they used the

cumulative number of changes to program elements (i.e., functions, types, and global variables) as

metrics for this law.

Alenezi and Almustafa [2015] conducted empirical investigation on OOS evolution by considering

five different sized and different domain OSS using Cyclomatic Complexity and SLOC as metrics.

They found support for applicability of Law of Continuing Change, Increasing Complexity and

Continuing Growth for OSS.

In this paper, aside function count metric used in previous studies, the quantification of the law of

Continuing Change is also measured using the following new metrics: parameters in functions,

returns in functions, statements in function body, and incremental differences in the following:

function count, parameters, returns and function statements.

RESEARCH METHOD

SAMPLE OSS (CASE STUDIES)

We downloaded the datasets which are used as case studies for this paper from UCR online

repository. The repository contains 9 merged source codes of evolutionary OSS written in C

language. The main reason for using a dataset placed on internet repository by a group of evolution

researchers and not trying to download all the publicly available official releases of the applications

that were used as case studies is that quantification of law of software evolution must be based on


37

empirical results, verifiable and repeatable, and made on a large scale, so that conclusions with

statistical significance can be achieved [Sjøberg et al.2008]. If software evolution is analysed with

data that is not available to third parties, it cannot be verified, repeated and replicated. It is

dangerous to build a theory on empirical studies that do not fulfil those requirements.

The datasets used as case studies in this paper are well evolved open source applications, namely:

Bison systems, Samba systems, SQLite Systems and Vsftpd systems. The above dataset are

selected because they have long term software evolution, sizable and actively maintained.

OVERVIEW OF THE APPLICATIONS USED AS CASE STUDY

Table 1 presents the sample OSS used for the empirical analysis.

Table 1: Sample OSS

OSS Description

Bison: Version

No. of Functions

No. of parameters

No. of returns

Statements

v1 v1.3 v1.5 v2.4.3 Bison is a general-purpose parser generator,

which can be use to develop a wide range of

language parsers, from those used in simple

desk calculators to complex programming

languages.

128 223 572 843

71 249 1003 1631

223 456 1062 1481

3646 7665 13214 19449

Samba: Version

No. of Functions

No. of Parameters

No. of returns

Statements

v1.5.14 v1.9.18 v2.2.12 v3.3.1 Samba is a tool suite that facilitates Windows-

UNIX/Linux interoperability. It is based on the

common client/server protocol of Server

Message Block (SMB) and Common Internet

File System (CIFS)

120 940 4001 13157

291 2071 10902 37292

379 2850 17142 69626

3039 25373 118597 506158

VSFTPD: Version

No. of Functions

No. of Parameters

No. of returns

v0.0.9 v1.1.0 V1.2.2 v2.1.0 Vsftpd stands for “Very Secure FTP Daemon”

and is the FTP server in major Linux

distributions like Ubuntu, Centos, Unix, Fedora

and the likes. It is licensed under the GNU

General Public License. It supports IPv6

325 401 450 590

414 641 724 939

461 564 719 997

http://searchnetworking.techtarget.com/definition/client-server

http://searchnetworking.techtarget.com/definition/protocol

http://searchcio-midmarket.techtarget.com/definition/SMB

https://en.wikipedia.org/wiki/GNU_General_Public_License

https://en.wikipedia.org/wiki/GNU_General_Public_License

https://en.wikipedia.org/wiki/IPv6


Statements 2987 4342 5470 7098

SQLite: Version

No. of Functions

No. of Parameters

No. of returns

Statements

v1.0.0 v2.0.7 v3.1.1 v3.6.11 SQLite is an implementation of a self-contained

SQL database engine. It is as a library in C

programming language. It is not a client-server

database management system and it comes with

a "shell" that can be used for command–line

interaction.

167 303 391 1284

393 677 886 3028

648 1197 1479 3556

6926 11527 13814 31126

Table 1 gives information on some selected versions of the sample OSS. We analysed the entire

versions of the sample OSS available on the UCR online repository as at the time of carrying out

this study. Bison has 33 versions (v1.0 - v2.4.3) released over a period of 22 years. 89 versions of

Samba released over a period of 15 years and grew from 5514 LOC to more than 1,000,000 LOC

were analysed. 60 versions of VSFTPD were analysed. 172 versions of SQLite were analysed;

starting with its first version v1.0 of 17723 LOC to v3.6.11 of 65108 LOC.

Measurement Tools and Metrics

Two static code analysing tools are selected namely Resource Standard Metrics (RSM) and

PMCCABE. They consume C/C++ and Java source code files and generate diverse metrics. RSM

tool was used to determine file function count, total function parameters, total function returns and

number of statements in functions.

PMCCABE is a command line tool in Linux environment, it calculates McCabe-style Cyclomatic

Complexity for C/C++ source code, statements in function, lines of code in function, blank lines

and the likes. PMCCABE uses per-function metrics and was run with Unix scripts in the Debian

command line. Metric values from PMCCABE tool was used for tracing individual functions and

their metrics values for analysis purpose. The metrics values of the two tools are correlated and

this gives confidence that the metrics values are correct; however, visual inspection was still done.

As regards Continuing Change law metrics, others including Neamtiu et al.[2013] have used

metrics like number of functions in each release, system size, module size, function modification,


39

function complexity, changes to types, global variables and the likes. Function count, parameters

in functions, returns of functions, statements in function body, incremental difference in the

following: function count, parameters, returns and statements in function are used in this study.

The above metrics have not been used in the previous studies except the function count metric.

The metrics are defined as follows.

i. Function Count (FC): this is the number of functions in a software version

ii. Functions Parameters (FP): this is the sum of all the parameters in functions of a software

version

iii. Functions Returns (FR): this is the sum of the return statements in functions of a software

version.

iv. Functions Statements (FS): the sum of executable expressions that terminates by semi-colon.

v. Incremental difference in the following: function count, parameters, returns and statements in

function.

vi. Function Count Incremental Difference (FC-ID): this is the difference between the sum of

functions in a system version and the immediate preceding one.

vii. Functions Parameters Incremental Difference (FP-ID): this is the difference between the sum

of parameters in functions of a software and its immediate preceding version.

viii. Functions Returns incremental difference (FR-ID): this is the difference between the sum of

returns in functions of a software and its immediate preceding version.

ix. Functions Statements Incremental Difference (FS-ID): this is the difference between the sum

of statements in functions of a software and its immediate preceding version.

Continuing change: E-type software systems must be continually adapted else they become

progressively less satisfactory. E-type software systems adapt to their environment and to

changing/increasing requirement from the users. The software systems considered as case studies

are widely used and well maintained, so they are good candidate softwares to investigate the law


of continuing change. Also, changes in kind of incremental difference in functions interfaces

(parameters and returns) and in the functions body like functions statements, depict continuing

changes occurring to the software systems in evolution. As the softwares are being continually

adapted to their environment, modifications to the function interfaces and body are inevitable.

Results and Discussion

Results

Having measured the sample OSS using the measurement tools, the following graphs depict the

relationship between the selected metrics under consideration. Figure 1 to Figure 9 depicts the

change behaviour of the four OSS. They show how incremental difference in Functions Count ,

Parameters, Returns and Statements changes in the four sample OSS.

Figure 1. Graph of Function Count Incremental Difference(FC-ID), Function Parameters

Incremental Difference(FP-ID), Function Returns Incremental Difference(FR-ID), Function

Statements Incremental Difference(FS-ID) against vsftpd pre versions of v1.0 on horizontal axis.

-50

0

50

100

150

200

250

300

0.0

.9

0.0

.12

0.0

.13

0.0

.14

0.0

.15

0.9

.0

0.9

.1

0.9

.2 p

re3

0.9

.2 p

re4

0.9

.2 p

re5

0.9

.3 p

re1

0.9

.3 p

re2

0.9

.3 p

re3

0.9

.3 p

re4

0.9

.4 p

re1

0.9

.4 p

re2

0.9

.4 p

re3

0.9

.4 p

re4

FC-ID

FP-ID

FR-ID

FS-ID

vsftpd


41

Figure 2a. Graph of Function Parameters Incremental Difference (FP-ID), Function Returns

Incremental Difference(FR-ID), function counts incremental difference(FC-ID) against vsftpd

version 1.0 releases on horizontal axis.

Figure 2b. Graph of Function Statements Incremental Difference (FS-ID) against vsftpd version

1.0 releases on horizontal axis.

-4

-3

-2

-1

0

1

2

3

4

1.0.2pre1 1.0.2pre2 1.0.2pre3 1.0.2pre4 1.0.2pre5

FC-ID

FP-ID

FR-ID

-60

-40

-20

0

20

40

60

1.0.2pre1 1.0.2pre2 1.0.2pre3 1.0.2pre4 1.0.2pre5

FS-ID

FS-ID

vsftpd

vsftpd


Figure 3. Graph of Function Counts Incremental Difference (FC-ID), Function Parameters

Incremental Difference(FP-ID), Function Returns Incremental Difference(FR-ID), Function

Statements Incremental Difference(FS-ID) against vsftpd version 1.1 releases on horizontal axis.

Figure 4. Graph of Function Counts Incremental Difference (FC-ID), Function Parameters

Incremental Difference(FP-ID), Function Returns Incremental difference(FR-ID),Function

Statements Incremental Difference(FS-ID) against vsftpd version 0.0.9 to version 2.1.0 on

horizontal axis.

-50

0

50

100

150

200

FC-ID

FP-ID

FR-ID

FS-ID

-200

0

200

400

600

800

1000

1200

0.0

.9

0.0

.14

0.9

.1

0.9

.2 p

re5

0.9

.3 p

re3

0.9

.4 p

re2

1.0

.2p

re1

1.0

.2p

re4

1.1

.1p

re1

1.1

.1p

re4

1.1

.3p

re1

1.1

.3r

1.2

.0p

re3

1.2

.0rc

2

1.2

.1p

re2

1.2

.2

2.0

.2 p

re1

2.0

.2r

2.0

.3r

2.0

.6

FC-ID

FP-ID

FR-ID

FS-ID

vsftpd

vsftpd


43

Figure 5a. Graph of Function Counts Incremental Difference(FC-ID), Function Parameters

Incremental Difference(FC-ID), Function Returns Incremental Difference(FR-ID) against bison

version 1.0 to version 2.4.3 on horizontal axis

Figure 5b. Graph of Function Statements Incremental Difference (FS-ID) against bison version

1.0 to version 2.4.3 on horizontal axis

-200

0

200

400

600

800

FC-ID

FP-ID

FR-ID

-4000

-2000

0

2000

4000

6000

8000

1

1.0

3

1.1

4

1.1

8

1.2

1

1.2

4

1.2

6

1.2

8

1.3

1.3

2

1.3

4

1.5

1.8

75

2.1

2.3

2.4

.1

2.4

.3

FS-ID

FS-ID

bison

bison


Figure 6a. Graph of Function Parameters Incremental Difference(FP-ID) ,Function Counts

Incremental Difference(FC-ID), Function Returns Incremental Difference(FR-ID) against samba

version 1.5.14 to version 3.3.1 on horizontal axis

Figure 6b. Graph of Function Statements Incremental Difference (FS-ID) against samba version

1.5.14 to version 3.3.1 on horizontal axis

-5000

0

5000

10000

15000

20000

25000

30000

1.5

.14

1.6

.07

1.9

.01

1.9

.06

1.9

.10

1.9

.14

1.9

.18

2.0

.3

2.0

.7

2.2

.1

2.2

.5

2.2

.9

3.0

.1

3.0

.5

3.0

.9

3.0

.13

3.0

.21

a

3.0

.25

3.0

.29

3.0

.33

3.2

.2

3.2

.6

3.3

.1

FP-ID

FC-ID

FR-ID

-100000

0

100000

200000

300000

1.5

.14

1.7

.07

1.9

.03

1.9

.09

1.9

.14

2.0

.0

2.0

.5

2.0

.10

2.2

.5

2.2

.10

3.0

.3

3.0

.8

3.0

.13

3.0

.22

3.0

.27

3.0

.32

3.2

.2

3.2

.7FS-ID

FS-ID

samba

samba


45

Figure 7a. Graph of Function Counts Incremental Difference(FC-ID), Function Parameters

Incremental Difference(FP-ID), Function Returns Incremental Difference(FR-ID) against sqlite

version 1.0.0 to version 3.6.11 on horizontal axis

Figure 7b. Graph of Function Statements Incremental Difference against sqlite version 1.0.0 to

version 3.6.11 on horizontal axis

-100

0

100

200

300

400

\sqlite-…

\sqlite-…

\sqlite-…

\sqlite-…

\sqlite-…

\sqlite-…

\sqlite-…

\sqlite-…

\sqlite-…

\sqlite-…

\sqlite-…

\sqlite-…

\sqlite-…

\sqlite-…

\sqlite-…

\sqlite-…

\sqlite-…

\sqlite-…

\sqlite-…

\sqlite-…

\sqlite-…

\sqlite-…

\sqlite-…

\sqlite-…

\sqlite-…

\sqlite-…

\sqlite-…

\sqlite-…

\sqlite-…

FC-ID

FP-ID

FR-ID

-1500

-1000

-500

0

500

1000

1500

2000

2500

\sq

lite

-1.0

.0.c

\sq

lite

-1.0

.10

.c\s

qli

te-1

.0.1

8.c

\sq

lite

-1.0

.25

.c\s

qli

te-1

.0.3

2.c

\sq

lite

-2.0

.6.c

\sq

lite

-2.1

.4.c

\sq

lite

-2.2

.3.c

\sq

lite

-2.4

.0.c

\sq

lite

-2.4

.7.c

\sq

lite

-2.5

.1.c

\sq

lite

-2.6

.1.c

\sq

lite

-2.7

.4.c

\sq

lite

-2.8

.4.c

\sq

lite

-2.8

.11

.c\s

qli

te-3

.0.3

.c\s

qli

te-3

.1.1

.c\s

qli

te-3

.2.1

.c\s

qli

te-3

.2.8

.c\s

qli

te-3

.3.6

.c\sqlite-…

\sq

lite

-3.4

.2.c

\sq

lite

-3.5

.6.c

\sq

lite

-3.6

.3.c

\sq

lite

-3.6

.8.c

FS-ID

FS-ID

sqlite

sqlite


Figure 8. Graph of Function Counts Incremental Difference(FC-ID), Function Parameters

Incremental Difference(FP-ID), Function Statements Incremental Difference(FS-ID), Function

Returns Incremental Difference(FR-ID) against sqlite version 2.1.0 to version 2.1.7 on

horizontal axis

Figure 9. Graph of Function Counts Incremental Difference(FC-ID), Function Returns

Incremental Difference(FR-ID), Function Parameters Incremental Difference(FP-ID), Function

Statements Incremental Difference(FS-ID) against sqlite version 2.4.0 to version 2.4.12 on

horizontal axis

Discussion

In Figures 1, it is seen in vsftpd systems evolution that in pre versions before the first main version

1.0, there were lot of early changes but close to the first major release candidate of v1.0 the changes

got vastly reduced as the trend becomes straight on the horizontal axis.

-10

0

10

20

30

40

50

FC-ID

FP-ID

FR-ID

FS-ID

-100

0

100

200

300

400

500

FC-ID

FP-ID

FR-ID

FS-ID

sqlite

sqlite


47

Figures 2a and Figure 2b show changes within vsftpd version 1.0 releases, it is noted that the

changes within stable branch of version 1.0 is within a short range i.e. no much changes; precisely

in Fig 2a, FC-ID the range is (1- (-1)) = 2; FP-ID is of range (3- (-3)) = 6; FR-ID is (1- (-1)) = 2;

FS-ID of Fig.2a is (53 - (-53)) = 106.

Figures 3a Show changes in function components of vsftpd v1.1 releases; it is seen that changes

within development branch of vsftpd v1.1 is more than that of vsftpd v1.0. The range of the

changes in the components/elements is wider i.e. the highest peak in FS-ID trend in v1.1 is 150

while its lowest trough is -30, the range is (150 - (-30)) = 180, while corresponding range of FS-

ID in v1.0 is (53 - (-53)) = 106, range of FC-ID in v1.1 is (12 - (-3)) = 15 while that of v1.0 is (1

- (-1)) = 2, range of FR-ID in v1.1 is (13 - (-4)) = 17 while that of v1.0 is (1-(-1)) = 2 ,this is same

with the trend of FP-ID. Also there are peaks and troughs in the Incremental Difference trends of

v1.1 than those of v1.0. However, there are continuing changes in both development branch and

stable branch.

In Figure4, it is shown that, more changes occur in the function statements than in the interfaces

of the function in term of parameters and return. Also, all trends from vsftpd version 0.0.9 to v2.1

is depicted, in contrast to vast reduction in changes prior or close to the first main version 1.0, in

the pre versions of a stable release which is to lead to the stable release, more changes do take

place close to the stable release candidate version. It is noted that the changes within the stable

branches are lower compared to the development branches, but there is continuing change in both.

The standard deviation of the incremental difference of statements in functions of vsftpd from

v0.0.9 to v2.1.0 is 161. The Skewness is 4.6 while Kurtosis is 27. Mean is 69.

The standard deviation of incremental difference of returns of vsftpd from version 0.0.9 to version

2.1.0 is 29, skewness is 5.99 and kurtosis is 40. Standard deviation is a measure of how widely

values are dispersed from average value, so standard deviation of 161 and 29 for incremental

difference of functions' statements and returns respectively shows that there is variance/dispersion

which depicts continuing change. Kurtosis characterizes the relative peakedness or flatness of a

distribution compared with the normal distribution. Positive kurtosis indicates a relatively peaked

distribution. Negative kurtosis indicates a relatively flat distribution. The incremental difference

of the statements and returns are of 27 and 40 respectively, which shows relative peakedness of


the metrics values distribution. Skewness characterizes the degree of asymmetry of a distribution

around its mean. Positive skewness indicates a distribution with an asymmetric tail extending

toward more positive values. Negative skewness indicates a distribution with an asymmetric tail

extending toward more negative values. The skewness of the incremental difference of statements

and returns in vsftpd systems are 4.6 and 5.9. The asymmetrical tails for both distributions extend

towards positive value. In all the descriptive statistics that have been considered for vsftpd version

0.0.9 to version 2.1.0 i.e. graphs, standard deviation, skewness and kurtosis; changes have been

depicted.

Figure 8 above shows graphs of Incremental difference for function counts, parameters, returns,

statement for sqlite v2.1 releases (development branch) .The changes in the development branch

of sqlite 2.1 is more than what happens in Figure 9 above for stable branch of sqlite 2.4, going by

the trends oscillations and variance, more continuing change occurred in sqlite v2.1 releases.

It can be seen in all the trends from Figure 1 to Figure 9, continuing change occurs to all the

systems in evolution. Also more functions and invariably more function components are added to

the systems as they are changing. Additions are more than deletions. In very large systems like

samba (there are versions with more than 1 million LOC), shrinkages/deletions occur less in

comparison to averagely large systems like sqlite systems. Changes within a major version

amongst the minor versions e.g. within major version 2.4 having minor versions v2.4.1, v2.4.2,

v2.4.3 and so on are less compare to changes that occur from major version to major regardless of

either development or stable major version i.e. from v2.4 to v2.5 to v3.0 and so on.

Incremental difference of samba (v1.5.14 to v3.6.11) functions statements, returns, parameters,

function counts are of mean 5717, 786, 420, 148 respectively and of standard deviation 23805,

3140, 1502, 533 respectively. This shows dispersion or variance which is invariably continuing

change. These statistical parameters values are also observed in other softwares of this project case

study. In the study of Neamtiu et al. [2013], cumulative number of changes to program elements

(i.e., functions, types, and global variables) of 9 evolutionary software systems were quantified


49

and continuing changes were found in all. Similarly, components of Linux Operating systems

(arch and drivers directories) were measured using number of files and Lines of Code metrics by

Israeli and Feitelson [2010], they also confirmed continuing changes occuring to in Linux

Operating system.

Stable major versions are systems like v1.0, 1.2, 2.2, 2.0, 2.4, 3.0, 3.4, 3.8 etc. Development major

versions have the second number in the versioning system to be odd number e.g. v2.1, 2.3, 3.5, 3.7

etc

CONCLUSION

In the pre versions, lot of changes do occur but when it is getting close to first release candidate of

version 1, the changes reduce but in stable branch after the first official release more changes do

occur close to the release of another stable or development major release.

Changes within stable branch is within a short range i.e. no much changes because simple

enhancements, updates, bug fixings are the major things that are usually done in the stable branch.

Changes within development branch are more than that of stable branch because many new small

files are created due to new features and functionalities in the development branch.

More changes occur in the function statements than in the interfaces of the function in term of

parameters and returns.

Changes within the stable branches are lower compared to the development branches, but there is

continuing change in both.

In all the trends, continuing change occurs to all the systems in evolution. Also more functions and

invariably more function components are added to the systems as they are changing. Additions are

more than deletions. In very large systems like samba (there are versions with more than 1million


LOC), shrinkages/deletions occur less in comparison to averagely large systems like sqlite

systems.

Changes within a major version amongst the minor versions e.g. within major version2.4 having

minor versions v2.4.1, v2.4.2, v2.4.3 and so on are less compare to changes that occur from major

version to major regardless of either development or stable major version i.e. from v2.4 to v2.5 to

v3.0 and so on.

Incremental difference of system components such as function statements, returns, parameters and

function counts show dispersion from their mean going by their standard deviation parameter. This

shows dispersion or variance which is invariably continuing change.

The systems under consideration observed versioning order as follows: stable major versions are

systems like v1.0, 1.2, 2.2, 2.0, 2.4, 3.0, 3.4, 3.8 e.t.c. Development major versions have the second

number in the versioning system to be odd number.

REFERENCES

ALENEZI, M., AND ALMUSTAFA, K. 2015. Empirical analysis of the complexity evolution in

open-source software systems. International Journal of Hybrid Information Technology, 8(2),

257-266.

EICK, S. G., GRAVES, T. L., KARR, A. F., MARRON, J. S., AND MOCKUS, A. 2001. Does

code decay? assessing the evidence from change management data. IEEE Transactions on

Software Engineering, 27(1), 1-12.

FERNANDEZ-RAMIL, J., LOZANO, A., WERMELINGER, M., AND CAPILUPPI, A. 2008.

Empirical studies of open source evolution. In Software evolution (pp. 263-288). Springer Berlin

Heidelberg.

GODFREY, M.W., & TU, Q. 2000. Evolution in open source software: A case study. In

Software Maintenance, 2000. Proceedings. International Conference on (pp. 131-142). IEEE.

HERRAIZ, I. 2008. A Statistical Examination of the Evolution and Properties of Libre

Software. PhD thesis, Universidad Rey Juan Carlos, Retrieved from

http://purl.org/net/who/iht/phd.


51

ISRAELI, A., AND FEITELSON, D. G. 2010. The Linux kernel as a case study in software

evolution. Journal of Systems and Software, 83(3), 485-501.

KOCH, S. 2007. Software evolution in open source projects—a large‐scale investigation.

Journal of Software: Evolution and Process, 19(6), 361-382.

LAWRENCE, M. J. 1982. An examination of evolution dynamics. In Proceedings of the 6th

international conference on Software engineering (pp. 188-196). IEEE Computer Society Press.

LEHMAN M. M. 1969. The Programming Process. IBM Res. Rep. RC 2722, IBM Res. Centre,

Yorktown Heights, NY 10594

LEHMAN, M. M. 1974. Programs, cities, students—limits to growth?. In Programming

Methodology (pp. 42-69). Springer New York.

LEHMAN, M.M. 1978. Laws of Program Evolution - Rules and Tools for Programming

Management, Proc. Infotech State of the Art Conf., Why Software Projects Fail,

pp. 11/1-11/25.

LEHMAN, M.M.1979. On understanding laws, evolution, and conservation in the large-

program life cycle. Journal of Systems and Software , 213 – 221.

LEHMAN, M. M. 1980. Programs, life cycles, and laws of software evolution. Proceedings

of the IEEE, 68(9), 1060-1076.

LEHMAN, M.M. 1985a. Program Evolution. In Program Evolution. Processes of Software

Change, M. M.Lehman and L. A. Belady (Eds.). Academic Press Professional, Inc.,

San Diego, CA, USA, 9–38.

LEHMAN, M.M. 1985b. The Programming Process. In Program Evolution. Processes of

Software Change, M. M. Lehman and L. A. Belady (Eds.). Academic Press Professional, Inc., San

Diego, CA, USA, 39–84.

LEHMAN, M.M. 1985c. Programs, Cities, Students: Limits to Growth? In Program Evolution

Processes of Software Change, M. M. Lehman and L. A. Belady (Eds.). Academic

Press Professional, Inc., San Diego, CA, USA, 133–164.

LEHMAN, M.M. 1989. Uncertainty in computer application and its control through the

engineering of software. Journal of Software Maintenance: Research and Practice, 1,1, 3–27.


LEHMAN, M.M. 1990. Uncertainty in Computer Application. Communications of ACM 33, 5,

584–586. Technical letter.

LEHMAN, M.M., PERRY, D.E., AND TURSKI, W. M. 1996.Why is it so hard to find Feedback

Control in Software Processes?, Invited Talk, Proc. of the 19th Australasian Comp. Sc. Conf.,

Melbourne, Australia, pp. 107-115.

LEHMAN, M.M. 1996b. Laws of Software Evolution Revisited. In Proceedings of the European

Workshop on Software Process Technology. Springer-Verlag, London, UK,

108–124.

NEAMTIU, I., XIE, G., & CHEN, J. 2013. Towards a better understanding of software evolution:

an empirical study on open‐source software. Journal of Software: Evolution and Process, 25(3),

193-218.

PIRZADA, S. S. 1998. A statistical examination of the evolution of the UNIX system (Doctoral

dissertation, Imperial College London (University of London)).

SJØBERG, D. I., DYBÅ, T., ANDA, B. C., &HANNAY, J. E. 2008.Building theories in software

engineering. In Guide to advanced empirical software engineering (pp. 312-336).

Springer London.

VASA, R. 2010. Growth and change dynamics in open source software systems.

Ph.D. Dissertation. Swinburne University of Technology, Melbourne, Australia.


53

Measuring the Impacts of E-Learning on Students’ Achievement in

Learning Process: An Experience from Tanzanian Public

Universities

Titus Tossy5

Mzumbe University, Tanzania

[email protected]

ABSTRACT

This paper is located within the global debates about the impact of e-learning as one of the ICT on

students’ achievements in teaching and learning process in universities. From the perspectives of

Tanzania, this paper provides a model for measuring the impact of e-learning on students’

achievements in universities. The rationale for the investigation stems from the notion that despite

the hundreds impact studies, the impacts of e-learning on student’s achievements remain difficult

to measure and open too much reasonable debate. This raised contradiction and elusive findings

on the conclusion based on the impacts of e-learning systems on student’s achievement. A Mixed

method research methodology involving survey and interviews was employed in the collection of

data for building the model. Multiple regressions technique was used to analyze the hypothesized

relationships conceptualized in the research model. The model was built and validated using

structural equation modeling and Delphi technique respectively. Measuring e-learning impact on

student’s achievements, indicators such as student engagement, student cognitive, performance

expectancy, student control, student satisfaction, continue using, student motivation, student self

esteem, student confidence on e-learning system have positive significance relationship with

students’ achievement. The model has the potential to policy makers, universities and other

5 Author’s Address: Titus Tossy, Mzumbe University, Tanzania "Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for

components of this work owned by others than IJCIR must be honored. Abstracting with credit is permitted. To copy otherwise, to republish, to

post on servers or to redistribute to lists, requires prior specific permission and/or a fee." © International Journal of Computing and ICT Research 2008.

International Journal of Computing and ICT Research, ISSN 1818-1139 (Print), ISSN 1996-1065 (Online), Vol.11, Issue 1, pp. 53 - 71, June

2017.



stakeholder to understand the impacts of e-learning after implementation in order to justify the

total investment based on that technology. The novelty of this research lies in the extension of the

findings in literature with constructs such as frequency use and intension to use e-learning in

learning context.

Categories and Subject Descriptors:

K.3.1 [Computers and Education] Computer Uses in Education;

General Terms: Collaborative Learning, Distance Learning

Additional Key Words and Phrases: E-learning, learning process, impacts of E-learning, Tanzania

Universities, Public Universities


Titus Tossy. Measuring the Impacts of E-Learning on Students’ Achievement in Learning

Process: An Experience from Tanzanian Public Universities, Vol. 11, Issue.1 pp 53 - 71.


1. INTRODUCTION

Information and Communication Technologies (ICTs) have influenced the landscape of education

sector by changing the way various education activities are being conducted. Rapid developments

of ICTs have improved access to and efficiency of teaching and learning processes in universities

(Lwoga and Komba, 2015), thereby leading to improved students’ achievements. This associated

academic career achievement provides the promise for meaningful employment of graduates as

well as movement towards a knowledge-based economy and rapid national economic growth

(Olson et al., 2011). Based on this reason, most governments and universities in developed

countries have invested in ICTs, e-learning systems in particular. As such, electronic learning

systems (e-learning systems) have become a major phenomenon in recent years (Tossy, 2012) as

transforms teacher-centered teaching and learning system into a student-centered one (Truncano,



55

2005). Further, this transformation enables students to develop their problem-solving abilities;

information reasoning and communication skills; improves creativity and other higher orderly-

thinking skills (Rosenblit et al., 2005). The system indeed changes the way in which teaching,

learning, and administration of education activities are conducted (Tossy, 2012; Lwoga and

Komba, 2015); offers efficient use of time and ease sharing of educational materials between

students and staff (Shivaraji et al., 2013) and improves the quality of teaching and learning (Kahiigi

et al., 2008; Jones, 2011).

Despite these notable attributes of utilization of e-learning in teaching and learning, its impact on

student’s achievements remain difficult to measure and open to debate as there are few conclusive

statements (Truncano, 2005; Rosenblit and Gros, 2011). Others further argue that there is a

contradiction on the conclusion on the impacts of e-learning systems on student’s achievement

(Hilz et al., 2001; Trancore, 2005). It is also argued that data to support the perceived benefits from

e-learning technologies are limited and evidence of effective impact is elusive (Eurydice, 2011;

Bocconi et al., 2013; Pandolfini, 2016). In developing countries, there is paucity of information

about the relationship between e-learning technologies and student’s achievement (Rosenblit et

al., 2011). There is thus a need to carry out more research, notably to develop useful indicators and

methodologies that need to be used in measuring the impact of e-learning in teaching and learning

in developing countries including Tanzania in order to guide policy formulation. This is important

because developing countries including Tanzania are still at very basic stage of e-learning

technology adoption. Tanzania needs to tap into experiences of universities in developed countries

that have long experience of using e-learning so as to formulate innovative corrective measures.

2. E-LEARNING

Wentling et al. (2000:5) define e-learning as:

““The acquisition and use of knowledge distributed and facilitated primarily by electronic means.

This form of learning currently depends on networks and computers but will likely evolve into

systems consisting of a variety of channels (e.g. Wireless, satellite), and technologies (e.g. Cellular

phones,etc.) as they are developed and adopted. E-learning can take the form of courses as well as


modules and smaller learning objects. E-learning may incorporate synchronous or asynchronous

access and may be distributed geographically with varied limits of time.” (Wentling et al., 2000:5).

E-learning captures a wide range of terms [Albert & Mori, 2001] referred to as ‘labels’ which have

been used to describe the concept of e-learning. These labels include, but are not limited to Web

Based Learning (WBL), Web Based Instruction (WBI), Web Based Training (WBT), Internet

Based Training (IBT), Online Resource Based Learning (ORBL), Advanced Distributed Learning

(ADL), Tele-Learning (T-L), Computer-Supported Collaborative Learning (CSCL), Mobile

Learning (M-learning or ML), Nomadic Learning, Off-Site Learning [Collis, 1996; Khana, 2005;

Yieke, 2005; Bates, 2001; Dam, 2004; Goodear et al., 2001; Pegler & Littlejohn, 2007; Dabbagh

et al., 2000; Barbara, 2002, 2004; Cramer et al., 2000; Salzbert & Polyson, 1995; Schreiber, et al.,

1998; Schank, 2001; Howard, 2003; and Singh, 2003]. The e-learning term is used interchangeably

with other related terms such as online learning, virtual learning, and web-based learning

(Twaakyondo, 2004).

While The use of e-learning has the added value of flexibility (” anywhere, anytime, anyplace”),

E-learning facilitates both learner engagement and the engaging of experiences (Uys, 2004;

Meyen, 2000; 2002). Meyen (2002) demonstrate how e-learning helps to overcome the traditional

barriers to education delivery. These barriers include lack of physical infrastructure, lack of

qualified teaching staff, absence of adequate education budgets, and the failure of traditional

pedagogy and curricula. East African countries are characterised by these barriers (Ndume et al,

2008). The failure of the government's efforts in building physical classrooms has created an

opportunity for innovative education delivery via e-learning (Yieke, 2005). As Alavi and Leidner

(2001) argues that e-learning’s importance will grow right across the educational spectrum from

primary to HEIs, the e-learning implementation in Tanzania HEIs is taking place despite the

various outlined barriers. The e-learning implementation differs from one HEI to another.


57

3. TANZANIA HIGHER EDUCATION STATUS

According to TCU (2010), the education sector in Tanzania has grown drastically for the past fifty

(50) years; this has been due to an increase in the number of Higher Education Institutions (HEIs).

The students’ enrolment has increased tremendously since independency. As MoEVT (2011)

states that the number of students enrolled in HEIs increased drastically. In 1961, Tanzania had

1,737 students enrolled in 4 HEIs, while in 2011 a total of 244,045 students in 358 HEIs (MoEVT,

2011). This emanated from free markets which encourages establishment of both private and

public HEIs, backed by various government policies on education sector such as Vision 2025, ICT

Policy and Higher Education Master Plan (HEMP), which enhance the establishment of both

private and public HEIs (Maliyamkono, 2006:396-445). Despite the fact that the number of HEIs

has increased since 1961, the pace of increase of students compared to overall national population

growth doesn’t match the enrolment offered by these institutions (Maliyamkono, 2006). This is

due to limitation on enrolment capacity, geographical constraints, cost of education, lack of enough

infrastructures, lack of qualified personnel and lack of innovative ideas (Chiemelie, 2012). In the

light of those challenges, e-learning is sought to be the ultimate solution in which the enrolment

does neither depend on the infrastructure nor geographical locations (Noe, 2005). As MoEVT

(2011) argues that the HEIs should deploy e-learning for their day to day training activities, in

order to minimize training cost and to remain competitive in the market. Furthermore, while MoCT

(2003) articulates the need for harnessing ICT opportunities to meet the vision 2025 goals by

blending strategic ICT leadership; ICT infrastructure; ICT Industry through Human Capital,

MoEVT (2007) stipulates that Tanzania needs national e-learning sensitization by stressing the

effort on applications such as distance education, e-learning, m-learning and blended learning.

4. E-LEARNING AT HEIS IN TANZANIA


Dr. Gajaraj Dhanarajan (2001:9), President of the Commonwealth of Learning, argued that:

“One would be foolish to question the importance of the internet and www for education in this

new decade; at worst, it has the ability to connect communities of learners and teachers and at its

best it could very well be the tool that education has been waiting for these past thousands of years;

its promise is only limited by the imagination and capacity of the people who can apply and benefit

from it”.

This kind of vision of a future electronically driven and inclusive education has been a driving

force for HEIs in Tanzania and has provided the spur to implement e-learning. As is the case with

other African countries, the rate of implementation of e-learning platforms in Tanzania is still very

slow despite the potential opportunities provided by open source technology and the conducive

environments created by the respective governments. There have been some initiatives on the part

of governments to develop ICT policies as a way forward in the implementation of e-learning. In

addition, there have been different round table conferences and the formation of the Tanzania

Commission of Universities (TCU) has fostered a debate on a common education delivery. For

example, Tanzania has abolished all taxes related to computers and related equipment and reduced

licence fees and royalties’ payable by the telecommunication operators (Morrison & Khan, 2003

and McPherson & Nunes, 2008). The more established public and private HEIs have managed to

implement e-learning platforms in Tanzania. They are implementing these using either open source

or customized platforms such as WEBCT, Blackboard, Moodle, Joomla, etc. Other universities in

the Tanzania have started the basic process of ICT infrastructure expansion to include local area

network implementation, Internet, computer labs and other facilities, as a way forward to the

establishment of e-learning (Sife, et al., 2007).

5. E-LEARNING MARKET AND THE DRIVERS OF CHANGE IN TANZANIA


59

While e-learning is not a new phenomenon in the developed world, it may be new to some

developing countries. Its market is rapidly increasing globally. While Merrill Lynch (2003) argues

that the e-learning is the fastest growing sector in the developed countries, many developing

countries (including Tanzania) are striving to implement e-learning in HEIs. Doughty et al. (2001)

and Saint (1999) have documented the rise of the virtual university in Africa (including Tanzania).

There are many e- learning initiatives in progress in Tanzania, such as Schoolnet, e-learning

centres, and African Virtual University (Ndume, et al., 2008; Sife et al., 2007). The increase in the

demand for higher education is one of the driving forces for implementing e-learning. Higher

population growth, lower education costs, increased access to education, and higher participation

rates in higher education changes the way firms organize work and cost-effectiveness and are

factors driving the implementing of e-learning in Tanzania (Ndume et al., 2008).

6. METHODOLOGY

Conceptual Model and Research Hypothesis Development

The research model for this study was formulated based on the concept of information system (IS)

success model adapted from DeLone and McLean (1992). The model is consisting of three

dimensions each consists three constructs as illustrated in Figure 1. This paper therefore uses this

conceptual model to underpin the measurement of the impact of e-learning system on student’s

achievement in Tanzania universities.

Independent/ Exogenous Variables Dependent/ Endogenous Variable


Figure 1: Conceptual Model Adapted from (DeLone and McLean, 1992)

Based on the conceptual model depicted in Figure 1, the following hypotheses were proposed:

Students’ acquisition of knowledge and skills (SACKS)

SACKS

SDMAL

SM

Student Engagement

Student Cognitive

Performance

Impact level

on E-learning

Students

Achievement

s

Student Control

Student satisfaction

Continue using

Student motivation

Student self esteem

Student confidence

Measurable indicators

sIndicatorsindicators

Latent Variables


61

H1. Students’ engagement on using the system has a significant positive relationship with their

achievements

H2. Students’ performance expectancy has a significant positive relationship with students’

achievement

H3. Cognitive learning using e-learning system has a significant positive relationship with

students’ achievement

Students’ development maturity as autonomous learner (SDMAL)

H4. Students control on using e-learning system has positive relationship with students’

achievement

H5. Students continue using e-learning system has positive relationship with students’

achievement

H6. Students’ satisfaction on e-learning system has positive relationship with students’

achievement

Students Motivation (SM)

H7. Student’s motivation on using e-learning system has positive relationship with students’

achievement

H8. Students self esteemed on e-learning system has positive relationship with students

achievement

H9. Students’ confidence on e-learning system has positive relationship with students’

achievement

The study used a survey design, involving 4 universities with long ICT experience. These were

thus purposively selected amongst 30 universities in Tanzania. Three hundred and fifty (350)

respondents used in this study, thereby 306 respondents equal to 87.5% representing the planned


respondent pool. The survey questionnaire consisted of five point Likert scales (Likert, 1932) was

employed. The in-depth interview was employed to collect qualitative data from ICT experts

during model validation. The data was then analyzed quantitatively and qualitatively respectively

to identify different indicators and aspects relating to the measure of the impact of using and not

using e-learning systems on students’ achievements. The empirical data were analyzed using

multiple regressions and structural equation modeling (SEM) using Statistical Package for Social

Science (SPSS). The multiple regressions were used in analyzing hypothesized relationships

conceptualized in the research model. In order to validate the model, the Delphi Technique was

employed (Harold and Murray, 1975) and a new model was developed accordingly (Rowe and

Wright, 1999).

7. RESULTS AND DISCUSSION

7.1 E-Learning Experience and Awareness

The study revealed that 75% of the respondents were exposed to e-learning systems based on

whether one had ever used it for learning; attended a course on e-learning (9.5%); heard about it

from a colleague of other institutions or seen a colleague using it (2%). It was further evident that

79% of students were aware of the use of e-learning frequently in their day-to-day learning

activities, while 65% were found to have intention of using e-learning methods in their academic

career. These results match with those of previous studies by Alexander (2008) and Mazman and

Usluel (2009) which found that the more a person is involved in Internet or Web activities, the

more they are likely to use e-learning. It is therefore more likely that, in developing countries

particularly Tanzania, use rate of e-learning methods is likely to increase if university can afford

to embrace them in institutional operations.

7.2 Indicators of the impact of e-learning

The results of the multiple regressions are shown in Tables 1, 2 and 3.


63

Table 1: SACKS indicators of Students’ achievements

Students

Achievement

(Measure)

Indicators

β

t-value

Significance

Tolerance

VIP

R2

SACKS

(Constant) .412 2.304 .012

.513

SE .268 .886 .271 .926 1.079

SC .618 7.854 .000 .641 1.560

PE .596 7.617 .000 .641 1.679

The results in Table 1 show that indicators such as student’s engagement (SE), student cognitive

learning using e-learning methods (SC) and the performance expectance (PE) on e-learning had

positive relationship with the student’s achievement.

Table 2: SDMAL indicators of Students’ achievements

Students

Achievement

(Measure)

Indicators

β

t-value

Significance

Tolerance

VIP

R2

SDMAL

(Constant) .412 2.304 .012

.684

SCO .191 .092 .244 .807 .931

SS .730 8.181 .000 .641 1.560

CU .592 6.211 .000 .641 1.559

The results [Table 2] further show that indicators such as students’ control on using e-learning

(SCO), students’ satisfaction (SS) and continued use of e-learning had positive relationship with

the students achievement.

Table 3: SM indicators of Students’ achievements


Students

Achievement

(Measure)

Indicators

β

t-value

Significance

Tolerance

VIP

R2

SM

(Constant) 1.106 6.88 .000

.896

SSE .323 4.409 .000 .641 1.560

MT .545 7.191 .000 .641 1.679

CON -.069 .881 .257 .903 1.108

Table 3 indicates that students’ self-esteem on using e-learning (SSE) and student motivation (SS)

had positive relationship with the students’ achievement with the exception of students’ confidence

on using e-learning.

7.3 A model for measuring e-learning impact on student achievement

The previously hypotheses were tested using SEM. Of the nine relationships, eight were

statistically significant (Table 4). These were student’s engagement (SS) (β = .268, p < .01);

performance expectance (β =.596, p < .01); student cognitive learning (SC) (β = .618, p < .01)

control on using e-learning (β = .191, p < .01); continued use of methods (β = .592, p < .01);

satisfactions (β = .730, p < .01); motivation (β = .545, p < .01); self-esteem (β = .323, p < .01) and

confidence on e-learning (β = -.069, p < .01). Only student confidence on using e-learning in

learning context was not supported.


65

Table 4: Summary of hypotheses tested

Hypotheses Accepted/Rejected

β, p <

.01

H1 Students’ engagement on using the system has a

significant positive relationship with their achievements Accepted .268

H2 Students’ performance expectancy has a significant

positive relationship with students’ achievement Accepted .596

H3 Cognitive learning using e-learning system has a

significant positive relationship with students’

achievement

Accepted .618

H4 H4. Students control on using e-learning system has


H5 H5. Students continue using e-learning system has


H6 Students’ satisfaction on e-learning system has positive

relationship with students’ achievement Accepted .730

H7 Student’s motivation on using e-learning system has

positive relationship on students’ achievement Accepted .545

H8 Students self esteemed on e-learning system has positive

relationship students’ achievement Accepted .323

H7 Students’ confidence on e-learning system has positive

relationship on students’ achievement Rejected -.069

With the latent variables presented in the conceptual model, Structural Equation Modeling (SEM)

approach (Bollen, 1998; Hoyle and Panter, 1995) was used to determine the cause-effect


relationships among the latent variables with their indicators and the e-learning on students’

achievement in education. Three regression models were developed and used to determine the

value of dependent variables. The regression models were developed for Students’ acquisition of

knowledge and skills (SACKS); Students’ development maturity as an autonomous learner

(SDMAL) and Motivation (SM). SACKS indicators were student engagement (SE); cognitive

capacity (SC) and Performance expectancy (PE). It was further apparent that SDMAL measurable

indicators were student control (SCO); satisfaction (SS); continued use (CU) and the measurable

indicators for SM were student motivation (MT); self-esteem (SSE) and confidence (CON).

Based on the findings, the initial regression models were as follows:

SACKS = 0.268SE + 0.596PE + 0.618SC R2 = 0.513…………………….. (1)

SDMAL = 0.191SCO + 0.592CU + 0.730SS R2 =0.684…………………. (2)

SM = 0.545MT + 0.323SSE - 0.069CON R2 = 0.896……………………… (3)

Where:

SE = Student Engagement: SC = Student Cognitive: PE = Performance expectancy

SCO = Student Control: SS = Student satisfaction: CU = Student Continue Using

CON = Confidence: MT = Student Motivation: SSE = Student Self Esteem

The entire model was found to have a significant fit for the study, as all the three regression models

had R2 > = 0.5 (Hoyle and Panter, 1995). All hypotheses from H1 up to H8 were found to have

significant positive relationship with the student’s achievement. However, on the hypothesis (H9),

the study revealed that students’ confidence on e-learning system had a negative relationship with

students’ achievement. However, this was contrary to the findings of the study conducted by Olson

et al., (2011).

Further from the findings above, it is clear that, student engagement, student cognitive capacity,

performance were the key indicators of the latent variable which is students’ acquisition of

knowledge and skills (SACKS) for one to realize how e-learning impacts on student teaching and

learning achievement. In addition students’ control, satisfaction and continued use of e-learning

strategies were indicators of the latent variable, which is Students’ development maturity as an


67

autonomous learner (SDMAL) which is known to have an influence on student’s teaching and

learning achievements. The findings further show that self esteem and motivation were indicators

of the latent variable which is Students Motivation (SM) that had positive significance on

students’ teaching and learning achievement. In exception the study shows that student’s

confidence on e-learning had a negative impact on student’s achievement. These findings agree

with those of Olson et al. (2011) and The McGraw Hill report (2011).

7.4 Model Validation

The model was validated using the Delphi Technique based on the assumptions that a group expert

judgment is better than an individual judgment (Amiresmaili et al., 2011). Therefore, two different

groups composed of panels of ICT experts were formed with the view to discuss and evaluate the

model. The experts were technical personnel; lecturers specialized in e-learning and consultants

of e-learning. All relevant determinant factors obtained from Section 2 were critically discussed

by panelists and compared. The expert judgments arising were then used to test the validity of the

model, which was then refined using inputs from the workshop. The model finally established was

a function of Students’ acquisition of knowledge and skills (SACKS), development maturity as an

autonomous learner (SDMAL), Motivation (SM) and Behavioral Intension (BI) as latent variables,

each with measurable variables as presented in section 3. This relation is depicted mathematically

as follows:

Measurement Model = f (SACKS, SDMAL, SM, BI) + e

This further shows that the model had the potential to improve the measurement of e-learning

impact on student’s achievement in order for the management at an institutional level to make

decision based on the impact. This is envisaged to help to realize the net benefit to justify the total

investment.

8. CONCLUSION AND RECOMMENDATION


This study shows that developed model [Figure 2] has the potential to be used in measuring the

impact of e-learning on students’ achievements in universities and other institutions. Results

obtained through a mixed research method approach revealed that Student Engagement (SE),

Cognitive capacity (SC), Performance expectancy (PE), Control (SCO), Continued use (CU),

satisfaction (SS), Confidence (CON), Motivation (MT), Self Esteem (SSE) are important

measurable indicators of the model. In particular, intention to use (IU) and the Frequency of using

(FU) e-learning are measurable variable from behavioral intension (BI) which are of particular

importance in evaluating its impact on students’ achievement. These are novel additions indicators

to measure e-learning technology utilization impacts using the developed model. These results call

for more research that focuses on evaluating the impact of e-learning systems on students’

achievement in teaching and learning using the developed model in this study. The developed

model as a result of this paper is important as it help policy makers, university managements and

other stakeholder to measure the impact of e-learning in order to understand the status of e-learning

for justifying the total investment in learning context.


69

Figure 2: A final model for Measuring the Impact of e-learning on Students Achievement

MT

SSE

SACKS

SDMAL

SM

Intention to Use (IU)

Frequency of Using

(FU)

E-

learning

Impact

on

Students

SE

SC

PE

CU

CON


9. REFERENCES

Alexander, B. (2008).‘‘Social networking in higher education’’, available at

http://net.educause.edu/ir/ library/pdf/PUB7202s.pdf: accessed on 13 November 2016.

Amiresmaili, M. et al. (2011). ‘‘A model for health services priority for Iran’’, Journal of

American Science, Vol. 7 No. 4.

Bocconi, S., Balanskat A., Kampylis P., & Punie Y. (Eds.).. (2013). Overview and analysis of

learning initiatives in Europe. Luxembourg: European Commission

Borgatti, S.P. & Cross, R. (2003), ‘‘A relational view of information seeking and learning in social

networks’’, Management Science, Vol. 49 No. 4, pp. 432-45.

DeLone, W. H., & McLean, E. R. (1992). Information Systems Success: The Quest for the

Dependent Variable. Information Systems Research, 3(1), 60–95

Eurydice. (2011). Key data on learning and innovation through ICT at school in Europe 2011.

Brussels: EACEA P9 Eurydice

Guri-Rosenblit, S., & Gros, B. (2016) E-Learning: Confusing Terminology, Research Gaps and

Inherent Challenges: International Journal of E-learning and Distance Education. Vol. 25, No. 1.

Hiltz, S. R., Zhang, Y., & Turoff, M. (2001). Studies of effectiveness of learning networks. Newark,

N.J.: New Jersey Institute of Technology

Hoyle, R.H., & Panter, A.T. (1995).Writing about Structural Equation Models, Sage, Thousand

Oaks, CA: pp 3-18

Harold, A.L. and Murray, T. (1975), The Delphi Method: Techniques and Applications, Addison-

Wesley, Reading, MA.

Jones, D. T. (2011). An Information Systems Design Theory for E-learning: A thesis submitted

for the degree of Doctor of Philosophy of The Australian National University. Pp-17-431

http://net.educause.edu/ir/


71

Kahiigi, E. K. et al “Exploring the e-Learning State of Art.” The Electronic Journal of e-

Learning Volume 6 Issue 2, pp77 -88, available online at www.ejel.org Electronic Journal e-

Learning Volume 6 Issue 2 2008 (77 - 88)

Lwoga, E. T & Komba M (2015). Antecedents of continued usage intentions of web based learning

management system in Tanzania: Education + Training, Vol. 57 Iss 7 pp. 738 – 756 Permanent

links to this document: http://dx.doi.org/10.1108/ET-02-2014-0014. Accessed on 23/3/2015.

Mazman, S.G. & Usluel, Y.K. (2009), ‘‘The usage of social networks in educational context’’,

World Academy of Science, Engineering and Technology, Vol. 49 No. 1.

Munguatosha, G.M. et al. (2011). A social networked learning adoption model for higher

education institutions in developing countries: On the Horizon, Vol. 19 Iss 4 pp. 307 –

320.available online at http://dx.doi.org/10.1108/10748121111179439. Accessed on 12/12/2016

Olsonurt, J., Tarkleson, E., Sinclair, J., Yook, S., & Egidio, R. (2011). An Analysis of e-Learning

Impacts & Best Practices in Developing Countries. With Reference to Secondary School

Education in Tanzania: pp. 1-53. Available online at http://tism.msu.edu/ict4d +1 517.355.8372.

accessed on 19/11/2016

Pandolfini, V. (2016). Exploring the Impact of ICTs in Education: Controversies and Challenges.

Italian Journal of Sociology of Education, 8(2), 28-53. doi: 10.14658/pupj-ijse-20

Rowe, G. & Wright, G. (1999), ‘‘The Delphi technique as a forecasting tool: issues and

analysis’’,International Journal of Forecasting, Vol. 15 No. 4.

Shivaraj, O. et al. (2013). Students’ Attitude towards the Uses of Internet: Indian Journal of

Library and Information Science, 7(1), 13-23.

Tossy, T. (2012). Cultivating Recognition: A Classic Grounded Theory of E-Learning Providers

Working in East Africa: pp.1-381. Available online at http://www.elearningcouncil.com. Accessed

on 2/5/2016

http://www.ejel.org/

http://dx.doi.org/10.1108/ET-02-2014-0014

http://dx.doi.org/10.1108/10748121111179439.%20Accessed%20on%2012/12/2016

http://www.elearningcouncil.com/


Trucano, M. (2005). Knowledge maps: ICTs in education. Washington D.C.: InfoDev, The

Information for Development Program.

Reducing Checkpoint Overhead in Grid Environment

Faki Ageebe Silas6, Jimoh Rasheed Gbenga

1,2Computer Science Department, Faculty of Communication and Information Science,

University of Ilorin, Ilorin-Nigeria

[email protected] ,

____________________________________________________________________________________________

ABSTRACT

Grid Computing has become major player in super-computing community. But due to the diversity

and disruptive nature of its resources, failure of jobs is not an exception. However, many

researchers have come up with models that enhance jobs survivability. Popular among this model

is checkpoint model which have the ability of saving already computed jobs on a stable secured

storage. This model avoids re-computing of already computed jobs from the scratch in case of

resources failure. But the time a job takes in checkpoinitng also becomes another task which adds

overheads to computing resources thereby reducing the resources performance. In order not to add

too many overheads to computing resources, the number of checkpoints must be minimized. This

6 Author’s Address: Faki Ageebe Silas6, Jimoh Rasheed Gbenga 1,2Computer Science Department, Faculty of Communication and Information Science, University of Ilorin, Ilorin-Nigeria [email protected] , "Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are

not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for components of this work owned by others than IJCIR must be honored. Abstracting with credit is permitted. To copy otherwise, to republish, to

post on servers or to redistribute to lists, requires prior specific permission and/or a fee." © International Journal of Computing and ICT Research 2008. International Journal of Computing and ICT Research, ISSN 1818-1139 (Print), ISSN 1996-1065 (Online), Vol.11, Issue 1 pp. 72 – 83, June

2017.




73

study proposed checkpoint interval models which is implemented based on fault index history of

computing resources. Failed jobs are re-allocated from their last saved checkpoint using an

exception handler. The study observed that arithmetic checkpoint model is better used when fault

index of computing resources is high while geometric checkpoint model is better when fault index

of resources is low.

Keywords: Arithmetic Checkpoint, Exception Handler, Fault Tolerance. Geometric Checkpoint


Faki Ageebe Silas7, Jimoh Rasheed Gbenga. Reducing Checkpoint Overhead in Grid

Environment, Vol. 11, Issue.1 pp 72 - 83. http://ijcir.mak.ac.ug/volume11-issue1/article5.pdf

1.0 INTRODUCTION

Grid has proven to be a great computational resource in computing circle. But due to its unstable

and unpredictable nature of its resources, resources and job failure is a norm rather than exception.

Systems and networks can fail, resources can be turned on and off and the introduction of more

7 Author’s Address: Faki Ageebe Silas7, Jimoh Rasheed Gbenga 1,2Computer Science Department, Faculty of Communication and Information Science, University of Ilorin, Ilorin-Nigeria [email protected] , "Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are


components of this work owned by others than IJCIR must be honored. Abstracting with credit is permitted. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee." © International Journal of Computing and ICT Research 2008.





users can result in resource starvations which significantly affect quality of service (QoS). To

maintain a tolerable QoS level for users, fault tolerance mechanisms are incorporated. Providing

fault tolerance in a grid environment while on the other hand optimizing resources utilization and

job execution time is a challenging task (Nakkeeran, 2015). This is because fault tolerance

measures increase computing overheads to grid resources. This study adopts checkpoint and

recovery system as a measure to increase job life line thereby optimizing job completion time. The

goal of this method is aimed at increasing the survivability of interrupted jobs by re-scheduling

them at last save checkpoint and minimizing the number of checkpoints in order not to incur too

many overheads. This model therefore, reduces the number of checkpoints and increase the

survivability of jobs thereby optimization resources performance (Li & Msacagni, 2003).

2.0 LITERATURE REVIEW

The desire of grid checkpoint service is to meet basic requirements which are: ability of software

to exchange information among resources, ability of grid middleware and infrastructures to

exchange and maintain vital information, and availability of checkpoint data to all resources

(Mangesh & Urmila, 2014). But achieving interoperability in grid is difficult due complexity and

unstable nature grid resources. The function of fault tolerance as opined by Paul and Jie (2003) is

to preserve the delivery of expected service despite the presence of faulty processors and declining

resources capability within the system. This means errors within grid system should be detected

and corrected promptly. Permanent fault that could cause great havocs should be located and

remove quickly. All these are to allow grid resources deliver a tangible and acceptable QoS. Due

to technological advancement and increase in autonomous systems, many researchers are

exploring ways to design and implement fault detective models that can predict and perform


75

recovery from crash processors of computing resources. These models aim at improving resources

performance and job survivability in the presence of failed processors, their effectiveness largely

depends on tuning runtime parameters such as the checkpoint interval and number of replicas

(Pankaj, 2011). Fault tolerance schemes in grid can either be pro-active or post-active (Garg &

Singh, 2011; Ganga & Karthik, 2012). In pro-active mechanism, the resources make a failure

prediction process before jobs are scheduled with the hope that it will follow the prediction process.

Whereas post-active mechanism handles the job failure after it has occurred. Most approaches

applied to fault tolerance in grid environment are on post-active rather than the pro-active

approaches (Sajjad & Babar, 2016)

According to Tanenbaun and Van Steen (2002), the most popular fault-tolerance model in use is

checkpointing which involved periodic saving of snapshot of job progress on a stable storage

device which can survive failures (hard disk). The information stored on the stable storage is called

a checkpoint. Any time a processor or job crashes, the last saved checkpoint is used to restart the

job or resource rather than from beginning. There exist varieties of checkpoint model but the

popular ones could either be categorized as coordinated, uncoordinated and communication

induced checkpointing (Elnozahy, Alvisi, Wang & Johnson, 2002).

The main advantage of checkpointing is that, it is a general technique which can be applied to any

type of parallel applications. The time to take a snapshot or make a checkpoint is also a task or job

which adds execution time overheads to a resource even when crashes have not occurred. This

overhead is dependent on the frequency at which checkpoints are taken and in turn depends on the

programmer or middleware designed.


To improve fault tolerance of grid system, Ndeye, Pierre and Ousmane (2003) proposed

hierarchical performance of checkpoint protocols grid computing which discussed protocols based

on rollback recovery that was classified into two categories: checkpoint based rollback recovery

and message logging protocols. However, the performance of protocols was observed to depend

on the characteristics of the system, network and applications running. In situations where the

computational intensity is low, the Algorithm Based on System checkpoint (ABSC) model is

applicable. Meanwhile, if the computational intensity is high, Algorithm based on application

Checkpoint (ABAC) model is more suitable though with production of slight overheads in fault

free situations and very reliable in faulty situations (Mangesh & Urmila, 2014). Adaptive fault

tolerant scheduling utilizes an adaptive number of job replicas according to the grid failure history.

This technique composed of Adaptive Job Replication (AJR) and Backup Resource Selection

(BRS) where AJR determines number of replicas according to selected resources.

To reduce checkpoint overheads, Ramesh (2016) proposed Optical Checkpoint Automation

(OCA). The goal of this model is to make the most effective use of grid resources and also to

improve throughput value in the mist of fault. This model focus on minimizing the effect of grid

faults and reduces fault recovery time using optimal automation of checkpoint. To evaluate the

performance of the model, fault index of resources were kept and analyzed and jobs were assigned

to resources based on next sequence of pattern of failure. The failure patterns were predicted by

using Hidden Markov Model (HMM) to assign checkpoint interval and also to provide automatic

failure replica (context file of checkpoint) to the grid resource. The proposed OCA model provide

a better performance as compare to adaptive algorithm though the model depends on previous

history of computing resources which implies that checkpoint interval of new resources are

unpredictable.


77

Checkpointing is the most common method to achieve fault tolerance. Though, there exist research

issues on how improve its efficiency and overheads can be, Sajadah (2011) proposed a novel

solution on how checkpoint can operate on parallel application in grid. The model allows

checkpoint on applications at regions where there no inter-process communication thereby

reducing the checkpoint overheads and checkpoint size. In Dilbag, Jaswinder, Singh, Amit and

Chhabra (2012), a novel technique to analyze the performance of checkpointing algorithms is

proposed. The model was implemented in a suitable cloud environment with six service node with

analysis made in terms of parallel jobs execution.

3.0 ARITHMETIC AND GEOMETRIC CHECKPOINT MODEL (AGCM)

This model allows users submit jobs to the grid scheduler. Each submitted job is then assigned to

a computing resource that matches the requirement by the scheduler as specified by user. As

computing resources continue execution of job, statistics about each job and their resources are

sent to the Grid Information Service (GIS) for storage. Resource capacity, checkpoint activities

and current load are some of the information’s that are stored in the GIS.

Scheduler

Schedule

Manager

Job

Dispatcher

Checkpoin

t Manager

Fault Index

Manager

User

Checkpoint

Server

Exception

Handler

GIS


Figure 1: Model with Checkpoint and exception handler.

Resources in grid environment are not of the same specifications. Specifications like processing

speed, internal scheduling policy and load factor varies for resources to resources. In the same

way, each job differs from other jobs by execution time, deadline, and so forth. Fault indexes here

indicate the frequencies of failure of a particular resource. This fault indexes are stored in the GIS

as performing history of every resources. The fault indexes are used by this model to determine

how frequently a checkpoint is made on a resource. This is to make checkpoint in a most reasonable

way bearing in mind that the time to make a checkpoint also adds to the execution time of

computing resources. In other to make checkpoint at appropriate and economical time, checkpoint

interval must be minimize. All computing resources with high fault index history use arithmetic

model (eqn I) to set their checkpoint while those resources with low fault index uses geometric

checkpoint model (eqn II) as shown in eqn I and II respectively.

𝐽endT = 𝐽startT + (𝑛 − 1) 𝐶𝑖 (1)

𝐽endT = 𝐽startT∗ 𝐶𝑖(n - 1) (2)

Where 𝐽endT is the end time of execution of job j

𝐽startT is the start time of execution of job j,

𝑛 is the number of checkpoint of job j with the execution time,


79

Ci is the checkpoint interval at which checkpoint is to be set.

The last saved checkpoint (𝐿𝑆𝑐𝑝) for any job as distributed to all computing resources can be

recovered by the model in eqn 3.

𝐿𝑆𝑐𝑝 = ∑ 𝐽𝑒𝑛𝑑𝑇𝑛𝑖=1 (3)

This shows that module(s) of a job can be recovered form one or more computing resources at the

time of crashes or processor failure.

Recovery and reassigning of failed jobs to new computing resources is achieved by an exception

handler.

Exception Handler Algorithm

Step 1: Start

Step 2: if every processor Pi is recovering from failure

Checkpoint is last storage saved massage for processor Pi

Step 3: for k = 1 to N (N is the number of processor holding the same jobs)

Do

For every processor pj that holds the same job

Step 4: invoke exception handler

Step 5: Stop

4.0 CHECKPOINT MODEL SIMULATION


A submitted job has an estimated start and end time with a checkpoint set a fixed interval in

between them. To test the performance of checkpoint overhead in the system, the study simulates

checkpoints based on our proposed model. Chosen any of the proposed models is based on fault

index of a computing resource. The performances of the two models are shown in figure 1. From

the simulation, job start time at first minute, and checkpoint interval is set every two minutes, eight

checkpoints were taken for each model.

Figure 1: Graph of checkpoint interval vs. time of execution

It can be seen that though, same numbers of checkpoints were set, execution time of jobs differs

as checkpoint progresses. At 4th checkpoint, model I (eqn I) has done 6 minutes while model II

(eqn 2) has done 8 minutes of execution. At 6th checkpoint, model I (eqn I) has executed for 10

minutes while model II (eqn 2) has executed for 32 minutes.

It can be observes that both models exhibit close related behavior up to the fourth checkpoint

interval, the execution time and the number of checkpoint made is almost the same. Though, the

numbers of checkpoints made are the same, the time used in execution of the job varies. Model II

0

20

40

60

80

100

120

140

1 2 3 4 5 6 7 8

Ex

ecu

tio

n T

ime

in

Se

con

ds

Number of Chekpoints

Model I

Model II


81

(eqn 2) execution time is more than model I (eqn I). The 8th checkpoint was when job has executed

for 14 minutes for model I (eqn I ) and 128 minutes for model II (eqn 2). This implies that Model

I will make more checkpoint than model II thereby incurring more checkpoint overhead. In terms

of recovery, model II will do more re-computing of failed job because the checkpoints are far apart

in the terms of time. Conclusively, model I is more suitable for computing resources that have high

fault index (mission critical jobs) and computing resources with poor uptime. Meanwhile Model

II is suitable for computing resources with more up time and less processor crashes.

The model recovery uses exception handler which is done from last checkpoint saved on stable

storage. The exception handling model handles fault and reallocate jobs automatically to other

resources without jobs beginning from the scratch because jobs start from the last served

checkpoint. The exception handler seamlessly relocates jobs to the scheduler making it less costly

and incurring little overhead in job.

5.0 CONCLUSION

In order to minimize checkpoint overheads incur by resources, the study proposed two models.

One was arithmetic (model I) and the other geometric (model II). Based on computing resource

history, a geometric checkpoint is good for a resource that has a good uptime time with less

withdrawer and crashes while arithmetic checkpoint is good for a mission critical job or resources

that is known to have a high faulty index. Exception handler was use to re-assigned jobs that were

unable to finishes due to processor crashes to new computing resources that meet their

requirement. Being a simple design, exception handler incurs little over heads and recovers jobs

in a seamless manner.


REFERENCES

DILBAG, S., JASWINDER, S. AND AMIT, C., 2012. Evaluating overheads of integrated

multilevel checkpointing algorithms in cloud computing environment, International. Journal of.

Computer Network and Information Security, 4(5), pp.29-38.

ELNOZAHY, E. A., ALVISI, L., WANG, Y.M. AND JOHNSON, D. B., 2002. A survey of Roll-

Back recovery protocols in massage -passing system, ACM Computing Survey, 34(3), pp.375-408.

GANGA K. AND KARTHIK S., 2012. A survey on fault tolerance in workflow management

and scheduling. IJARCET, 1( 8) , p.176.

GARG, R. AND SINGH, A.K, 2011. Fault tolerance in grid computing: state of the art and open

issues. International Journal Computer Science Engineering Survey.2(1), p.88. doi:

10.5121/ijcses.2011.2107.

LI, Y., AND MSACAGNI, M. 2003. Improving performance via computational replica on a large

scale computational grid. Proceedings of third International Symposium on Cluster Computing

and the Grid, Tokyo-Japan.

MANGESH, B. AND URMILA, S. 2014. Checkpointing based fault tolerant job scheduling

system for computational grid. International Journal of Advancements in Technology, 5(2).


83

NAKKEERAN, M. 2015. A survey on task checkpointing and replication based fault tolerance in

grid computing. International Research Journal of Engineering and Technology (IRJET), 3(9),

pp.832-838.

NDEYE, M. N., PIERRE S. AND OUSMANE, T. 2003. Performance comparison of hierarchical

checkpoint protocols grid computing. International Journal of Interactive Multimedia and

Artificial Intelligence, 1(5), pp.46-53.

PANKAJ, G. 2011. Grid computing and checkpoint approach. International Journal of Computer

and management Studies. 11(1).

PAUL, T. AND JIE, X. 2003. Fault tolerance within a grid environment, University of Durham

DHI 3LE, United Kingdom.

RAMESH, B. 2016. An optimal checkpoint automation mechanism for fault tolerance in

computational grid. International Journal of Scientific & Engineering Research, 7(2), pp.1012-

1019.

SAJADAH, K. 2011. Checkpointing of parallel applications in a grid environment, A dissertation

submitted to the University of Westminster for the degree of Master of Philosophy, Centre for

Parallel Computing, School of Electronics and Computer Science, University of Westminster,

London, United Kingdom.

SAJJAD, H. AND BABAR, N. 2016. Fault tolerance in computational grids: perspectives,

challenges, and issues, SpringerPlus, 5(1), Nov 18. doi: 10.1186/s40064-016-3669-0.

TANENBAUN, A.S. AND VAN STEEN, M. 2002. Distributed Systems, Principles and

https://dx.doi.org/10.1186%2Fs40064-016-3669-0


Paradigms, Prentice Hall

Keystroke Dynamics Authentication for a Web-Based Sales and

Stock Solution

Oluwakemi C. Abikoye8 and Bilikis T. Sanni

Department of Computer Science, University of Ilorin, Ilorin, Nigeria

*Corresponding author: [email protected]

ABSTRACT

The importance of security in any financial sector cannot be over emphasized, a single factor

authentication (username and password) is no longer sufficient. Therefore, a multifactor

authentication is required to address security in financial systems. In this work, keystroke

dynamics, a new behavioral biometric is proposed as an additional security measures. Keystroke

dynamics is a very important tool for authentication with a high level of security which can be

implemented to enhance other forms of authentication. This is achieved using static keystroke

approach on passwords, statistical feature extraction method to analyze the keystroke time from

hold-time and latency, and direct comparison with thresholds based on password length for users

verification and authorization. This system is used in a web-based small and medium scale

8 Author’s Address: Oluwakemi C. Abikoye8 and Bilikis T. Sanni, Department of Computer Science, University of Ilorin, Ilorin, Nigeria

Corresponding author: [email protected] "Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are


components of this work owned by others than IJCIR must be honored. Abstracting with credit is permitted. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee."

© International Journal of Computing and ICT Research 2008.





85

enterprises (SMEs) sales and stock solution and also it provides a stronger security measure than

the conventional password authentication.

Keywords: Keystroke Dynamics, Password, Security, Multifactor Authentication, Biometric,

Small and Medium Scale Enterprises (SMEs)

IJCIR Reference Format: Oluwakemi C. Abikoye and Bilikis T. Sanni. Keystroke Dynamics

Authentication for a Web-Based Sales and Stock Solution, Vol. 11, Issue.1 pp 84 - 113.


1. INTRODUCTION

In this research work, an improved small and medium scale enterprise in terms of security is

developed using keystroke dynamics biometrics to enhance the usual password authentication.

Keystroke dynamics authentication is a multifactor authentication based on individual typing style

that can be used with other forms of existing security biometrics that involves typing like the

password used in this work. Keystroke dynamics is a very secure behavioral biometric, since each

individual have different typing styles (Sawant, Nagargoje, Bora, Shelke, and orate, 2013). The

purpose of this behavioral biometrics is to serve as a two-step security verification for user logins

to facilitate higher level of security during verification and authentication so that only authorized

users will have access to the information of the software.

Authentication is the process of identifying and verifying an individual for security purpose. User

authentication is classified into three main categories (Modi, Upadhyay, and Thakor 2014;

Abualgasim, and Osman 2011) which include; knowledge, object or token and biometrics based

authentication. Knowledge based authentication is a type of authentication that is based on

something one knows and is characterized by secrecy. e.g. Personal Identification Number(PIN)

codes and passwords. PINS and passwords are individual user unique set of alphabets, numbers,

or alphanumeric used for security during registration usually with a username.



Figure 1: Classification of Authentication

Token based authentication is a method of authentication that is based on something one possess

such as keycard or badge while Biometrics based authentication is a type of authentication that is

based on individual physical or behavioral features.

Biometrics is derived from two Greek words bio meaning “life” and metric meaning “to measure.

In computing, biometrics techniques are mainly used for user authentication for measuring and

Knowledge based

Authentication

Physiological

biometrics

KEYSTROKE DYNAMICS,

gait, voice, lip movement

and signature

Retina scanning, voice

recognition, fingerprints,

face recognition, palm-print

Behavioral

biometrics

Biometric based Object or token

based


83

analyzing individual’s physiological or behavioral features like keystroke, gait, fingerprints,

signature, lip movement, voice patterns, typing patterns, facial patterns, and palm-print for the

identification and verification purpose.

1.1 Types of Biometrics

There are two main categories of biometric based authentication which includes:

Physiological biometrics

Behavioral biometrics

Physiological Biometrics

Physiological biometrics is a type of biometric that measures individual physical features of a

certain part of the body. e.g. human eye (this involves iris and retina scanning), voice recognition,

fingerprints, face recognition, palm-print etc. Physiological biometrics is normally considered

more effective than behavioral biometrics because they are fairly consistent over time and are

unique, but are very expensive and require different equipment.

Behavioral Biometrics

Behavioural biometrics is a type of biometrics based on what a person does, or how the person

uses the body. e.g. keystroke dynamics, gait, and signature. Behavioural biometrics is considered

weaker when compared to the physiological biometrics because they are less constant and can vary

over time. Although Behavioural biometrics have unstable qualities, they are still successfully used

because it is still difficult to imitate others behaviour such as typing rhythm. Keystroke dynamics

is one of the cheapest biometric methods to implement since the only hardware required is the

keyboard.


1.2 Keystroke Dynamics

Keystroke dynamics is one of the several innovative technologies used to automate the process of

authenticating an individual based upon a unique and personal behaviour. Keystroke dynamics

involve a technique of analyzing how a user types at a terminal by observing the keyboard in order

to find the user based on his or her usual typing rhythm features (Sawant et al. 2013). Keystroke

dynamics also involve inspecting timing features of individual’s typing in order to identify patterns

in their keystroke data. This may sometimes include the analysis of features such as the pressing

time, releasing time and the latency of all authorized users extracted while typing their password

during registration process.

Most of the papers used statistical methods and neural networks for keystroke based

Authentication. Collection of data needs keyboard and software (Sawant et al. 2013). All

Keystroke mechanism depends upon the key press and key release duration. The time

duration between key press and key release convert into security parameters (Singh, and

Arya, 2010)

1.3 Small Medium Enterprises

There have been related studies in the past and still on-going by different researchers, reviews have

shown that integrated business software covers all relevant aspects of an enterprise (Davenport,

1998). The Small and Medium Scale Enterprises (SMEs) provide the cornerstones on which any

country’s economic growth and stability rest. Based on this, there is a need to take all available

security measure on such systems. This work uses keystroke dynamics biometrics authentication

as a plus to the usual login process. Research by experts has recommended and suggested

principles that modern enterprise systems should be service-oriented, constructed in a modular

form and communicates via Web services (Schmitt, 2007). This keystroke dynamics authentication

SMEs research work is cost effective which helps to solve the problem further analyzed by Scott

& Vessey (2002). Winkelmann and Klose (2008) stated that in small and medium companies in

particular, the introduction of a new system or the migration of old to new platforms is

accompanied by high costs in relation to turnover and naturally by associated risks because


85

integrating keystroke dynamics authentication to such system attracts no additional cost. It’s

almost free in terms of hardware because the most significant device used for data input is the

keyboard which is part of the existing hardware of the system.

2. RELATED WORK

Keystroke dynamics authentication, a behavioral biometrics has drawn a lot of research interest

and a number of techniques in different aspects were proposed. Several methods and modules have

been suggested with special emphasis on data collection, feature extraction, storage database and

matching process. Most of the papers used statistical methods and neural networks for keystroke

based Authentication, collection of data needs keyboard and simple software (Sawant et al. 2013).

Avasthi and Sanwal (2016) carried out a comprehensive study on Keystroke Dynamics based on

identification, verification, methods and performance measures & metrics. Modi, Upadhyay, and

Thakor (2014) compared existing systems based on some features such as the authentication

provided, classification method, outlier handling, keystroke hardening scalable, False Acceptance

Rate (FAR), False Rejection Rate(FRR) ration, limit ID and password text.

Giot, El-Abed, and Rosenberger (2014) used two (2) main families of keystroke dynamic methods

in making use of statistical methods; (1) The static families, where the user is asked to type several

times the same string in order to build its model, and (2) The dynamic families which allows the

authentication of individuals independently of what they are typing on the keyboard.

In Sawant et al. (2013) proposed two algorithms to implement the keystroke efficiently. The

Gaussian Probability Density Function and Direction Similarity Measures were proposed. The

fusion of the two algorithms was done by different methods which from among them the AND

rule showed the best result. The resulted mean, standard deviation and weight formula used to

calculate the weight in the first step and the login time keystroke data was compared with the

registered mean ± standard deviation which resulted into match count in the second step. Monte

Carlo approach was used for Data Collection; the time duration between key press and key release

was converted into security parameters while the Parallel Decision Tree (DT) was used for the

identification of a genuine user.


Shukla and Solanki (2013) focused on the problem of human emotion recognition in the case of

naturalistic. Users’ behavioral or physiological patterns were mapped to emotional categories

collected. Six basic emotional classes were worked on, confidence, sadness, nervousness,

happiness, tiredness, hesitation. The study aimed to develop a web application that recognizes

human emotional states using four modules; Data Collection Process which consists of gathering

and labeling users’ keystroke and mouse data. The feature extraction and attribute selection which

reduces the number of attribute to facilitate the classification process were performed. Data

Labeling and the data provided by the human were labeled while questioning the best accuracy of

the trained system. Classification: fuzzy logic was used as a solution to recognize user’s emotional

state.

Bajaj and Kaur (2013) applied statistical method to measure the mean time and average time and

developed an application using java. Some studies based on selection process on normality

statistics applied. Abernethy and Rai (2012) captured time values associated with keystroke events,

keystroke duration and digraph latency metrics were calculated from the raw data files and

Artificial Neural Networks were used for classification, and results were calculated as the false

acceptance rate (FAR) and the false rejection rate (FRR). Karnan and Krishnaraj (2012)

implemented statistical method on various keystroke dynamics feature extracted: (i) Duration

(Amount of a time a key is pressed), (ii) Latency (Differences of time between two key events),

(iii) Mean, standard deviation (Mean and standard deviation value of each type of PIN), (iv) Press–

Release (Latency between pressing and releasing the key), (v) Digraph

Al-Jarrah (2012) used a password typing rhythm to detect the actual and unauthenticated user. The

study introduced two phases; training phase and testing phase. The calculation of key down/up

time of every key and latency time between keys were done. The comparison in testing phase

check the value in binary for analysis, the binary score is 1 if it is within a close distance threshold

from the median stored else 0. The use of benchmark data in the study of the detection performance

of authentication systems is a scientific approach. A novel technique was proposed for free text

keystroke dynamics in Singh & Arya (2011), keys were classified into two halves (left-right) and


87

four lie (total eight groups). They used timing vector to calculate the flight time. Timing vector

technique used to define the difference between fraud and actual users. The best results were

obtained and very supporting.

In Giot et al. (2009) a comparative study was conducted, considering the operational constraints

of collaborative systems. The main scheme used includes:(i) the enrollment (which consists of

registering the user in the system) embeds the data capture, eventually some data filtering, the

feature extraction, and the learning step; and (ii) the verification process realizes the data capture,

the feature extraction and the comparison with the biometric model. Depending on the studies, this

scheme may be slightly modified. Some algorithms can also adapt the model of the user by adding

the template created during the last successful authentication.

3. METHODOLOGY

Nigeria is still a developing country especially when it comes to technology, and majority

of the enterprise in Nigeria especially the small and medium scale enterprise still operate on the

manual method of inventory. The staff record, stock record and daily sales are still being recorded

manually especially in the small enterprises because of the fear of expansion. In this research, after

series of direct interview, it was deduced that two things causes the fear, inadequate knowledge

about computing technology and assumed the new technology is too expensive. This project,

design with the aid of computer system, an inexpensive solution that doesn’t require expert

knowledge to effectively handles operations of the SMEs in Nigeria from the information gathered

by visiting some stores (small and medium enterprises) and designing a web-based application

using PHP, JavaScript and HTML at the front-end, MYSQL at the back-end for the database and

also apply keystroke dynamics biometrics on the password using PHP, JavaScript ( Query) and

HTML at the front-end, MYSQL at the back-end for the database, to be able to give the

organization (SMEs) a more secure system than using password alone. The cost of adding

keystroke dynamics to any system is almost free since the only hardware required is the keyboard

which is already part of the Computer system. This keystroke dynamics authentication security

based system is designed to eradicate the problems associated with the old system of unauthorized

users accessing the data of an organization to perform any form of fraudulent actions.


In this new system, though if the unauthorized users lay hands on password (that is guessed,

lost or stolen), the imposter would not be able to gain access to the system. This is so because

during the registration process of new system users, as the password is being typed, keystroke time

is also been calculated for every users. Even if unauthorized user knows the password, he/she

would not have access to the new system because of the individual keystroke time access control.

To access the new system, the inputted username, password and keystroke time must be the same

in the database

3.1 Basic Modules Used

There are four main basic modules used in keystroke dynamics from the papers reviewed Modi,

Upadhyay, and Thakor (2014); Sawant, Nagargoje, Bora, Shelve, and Borate (2013); Bajaj, and

Kaur (2013); Abernethy, and Rai (2012), which includes:

Data Collection

Feature Extraction

Storage Database

Matching Process

3.1.1 Collection of Data

This is the enrollment phase where by the authorized staff (both the admin staff and the

sales representative) register to create an account in the sales and stock web application. During

registration, each user registers with their specific password and username which will be saved in

the database against that specific user and while processing the password, the keystroke dynamics

program will also be performing its operation by using Static Approach where the same username

and password are maintained for each authorized users in order to build its model. In Static

approach, the system checks the user only at the authentication time. It provides additional security

than the username/password. It also provides more robust user verification than simple passwords.

In this approach, the analysis is performed on the password while typing during registration for all

the individuals using the application. The minimum password length used is 4 while the maximum

is 30. The static analysis is done at login time in conjunction with the password authentication


89

methods. The username and password used during registration are saved against each user for

verification during login. Collection of data is done using keyboard and software.

3.1.2 Feature Extraction

The keystroke data is based on the password using the system time. The key press (the time

in which the key is held down), key release (the time in which the key is released), the latency (the

time between two consecutive keys) and the password length (the total keys of the password) of

each users (staff) which are produced while typing each characters of user password for the

registration. The passwords are further analyzed using the statistical method to find the keystroke

time which will be stored in the database for future verification for each registered users.

Latency is the distance between two successive keys.

𝐿 = 𝐾2𝑃 − 𝐾1𝑅

Where 𝐿 = Latency, 𝐾1𝑅 = the time Key 1 is Released, and 𝐾2𝑃 = the time Key 2 is Pressed.

Duration is the amount of time a key is pressed or the hold-time for each key of individual

password. The difference between the time a key is released and pressed.

𝐷 = 𝐾1𝑅 − 𝐾1𝑃

Where 𝐷 = Duration, 𝐾1𝑅 = the time Key 1 is Released, and 𝐾1𝑃 = the time Key 1 is Pressed.

Digraph is the duration or hold-time of a key plus latency between successive keys.

𝐷𝐼 = 𝐷𝐾1 + 𝐿𝐾1,2

Where 𝐷𝐼 = Digraph, 𝐷 = Duration, 𝐾1 = Key 1, 𝐿 = Latency, and 𝐾1,2 =Between Key 1 and

Key 2.

Keystroke Time is the amount of time of typing the whole password used considering the key press

time, key release time, latency and password length.

Password Length is total keys of the password.

𝐾𝑇 =𝐷𝐼𝐾1+𝐷𝐼𝐾2+⋯+𝐷𝐼𝐾𝑛

𝑃𝐿

Where 𝐾𝑇 =KeystrokeTime, 𝐷𝐼 = Digraph, 𝐾1 = Key 1, 𝐾2 = Key 2, and 𝐾𝑛 = Key n, and

𝑃𝐿 = Password Length


Figure 2: Feature Extraction

3.1.3 Storage Database

The database stores all information of the new system in a systematic way, which makes

the update, retrieval and maintenance of the new system easy. After featured extraction and the

statistical calculations on the feature, the keystroke time is stored on the database which will be

retrieved for login verifications. The database houses all the business information; inventories of

staff, stock and daily sales (name of goods, quantity of goods sold, price of each item sold (unit

KEY 2 KEY 1

LATENCY

DURATION/

HOLD-TIME DURATION/

HOLD-TIME

PRESS

EVENT

RELEASE

EVENT

PRESS

EVENT

RELEASE

DIGRAPH


91

price) and the cost of goods sold which is its unit price multiplied its quantity, against each sales

representative).

3.1.4 Matching Process

In the matching process, genuine users are identified by direct comparison. During the

login process, the system compares the user name and password to the pre-registered data on the

database; if it doesn’t correspond it pops-up incorrect username or password with a back check

box to restart the process, otherwise as the user is typing his password to login, the keystroke also

counter check. The keystroke time of the present login is compared with the existing keystroke

time stored in the database. This system uses threshold considering the password length. The

default threshold is ±10 for the keystroke time (in microseconds) for password length that ranges

from 4:10 characters, for password length that ranges from 11:20 characters, threshold ±15 is set

for the keystroke time (in microseconds) and for password length that ranges from 21:30

characters, threshold ±20 is set for the keystroke time (in microseconds). These thresholds are

used during the verification between the present keystroke time and saved keystroke time for

authentication. If the present keystroke time falls between the given range, the user is able to login

otherwise pops up accessed denied which directs the user back to the login page to enter the

password again and follow the same process.

3.2 Proposed Algorithm

3.2.1 Algorithm for Creating New User

Step1: User opens the create account page to fill

Step 2: Input username, password, retype password, and full name

input name="username" type="text" id="username" size="30"

input name="pass1" type="password" id="pass1" size="30"

input name="pass2" type="password" id="pass2" size="30" and

input name="name" type="text" id="fullname" size="30"

Step 3: Select branch and role from branch and role table respectively in the database

Step 4: Submit


Step 5: The System Confirm password and the retyped password to be the same

if (password == retyped password)

{Create Keystroke Dynamics for Password}

Else {"Your Password do not match, Please try again!!!"}

Step 6: The system Extract keystroke feature from the password to create keystroke time

The key press (the time in which the key is held down) is extracted

The key release (the time in which the key is released) is extracted

The latency (the time between two consecutive keys) is extracted and

The password length (the total keys of the password) is extracted

The keystroke time is calculated using statistical method.

Step 7: Create and Save Authorized user username, password and Keystroke time in database

against individual username.

Step 8: display account creation page

3.2.2 Algorithm for Feature Extraction

Begin;

Input password; //Character by character

𝒊𝒏𝒕 𝑆𝑖𝑧𝑒 = 𝑠𝑡𝑟𝑔𝑙𝑒𝑛(𝑝𝑎𝑠𝑠𝑤𝑜𝑟𝑑);

𝒊𝒇 𝑖 = 0; //extract key press & key release

𝒅𝒐 { 𝒇𝒍𝒐𝒂𝒕 𝑡𝑖𝑚𝑒𝑝𝑟𝑒𝑠𝑠[𝑠𝑖𝑧𝑒], 𝑡𝑖𝑚𝑒𝑟𝑒𝑙𝑒𝑎𝑠𝑒[𝑠𝑖𝑧𝑒];

𝑡𝑖𝑚𝑒𝑝𝑟𝑒𝑠𝑠[𝑖] = 𝑡𝑖𝑚𝑒𝑠𝑡𝑎𝑚𝑝(𝑒𝑣𝑡. 𝑘𝑒𝑦𝑝𝑟𝑒𝑠𝑠); 𝑡𝑖𝑚𝑒𝑟𝑒𝑙𝑒𝑎𝑠𝑒𝑠[𝑖]

= 𝑖𝑚𝑒𝑠𝑡𝑎𝑚𝑝(𝑒𝑣𝑡. 𝑘𝑒𝑦𝑟𝑒𝑙𝑒𝑎𝑠𝑒);

𝑖 + +; }

𝑤ℎ𝑖𝑙𝑒 (𝑖 < 𝑠𝑖𝑧𝑒);

// Calculate duration, latency &digraph.

𝒇𝒐𝒓 (𝑗 = 0; 𝑗 < 𝑠𝑖𝑧𝑒; 𝑗 + +; )

{ 𝒇𝒍𝒐𝒂𝒕 𝑑𝑢𝑟𝑎𝑡𝑖𝑜𝑛[𝑠𝑖𝑧𝑒], 𝑙𝑎𝑡𝑒𝑛𝑐𝑦[𝑠𝑖𝑧𝑒], 𝑑𝑖𝑔𝑟𝑎𝑜ℎ[𝑠𝑖𝑧𝑒];

𝑑𝑢𝑟𝑎𝑡𝑖𝑜𝑛[𝑗] = 𝑡𝑖𝑚𝑒𝑟𝑒𝑙𝑒𝑎𝑠𝑒[𝑗] − 𝑡𝑖𝑚𝑒𝑝𝑟𝑒𝑠𝑠[𝑗];


93

𝑙𝑎𝑡𝑒𝑛𝑐𝑦[𝑗] = 𝑡𝑖𝑚𝑒𝑟𝑒𝑙𝑒𝑎𝑠𝑒[𝑗 + 1] − 𝑡𝑖𝑚𝑒𝑝𝑟𝑒𝑠𝑠[𝑗];

𝑑𝑖𝑔𝑟𝑎𝑝ℎ[𝑗] = 𝑑𝑢𝑟𝑎𝑡𝑖𝑜𝑛[𝑗] − 𝑙𝑎𝑡𝑒𝑛𝑐𝑦[𝑗]; }

//Calculate keystroke as the mean value of the digraph

𝒇𝒐𝒓 (𝑖 = 0, 𝑖 < 𝑑𝑖𝑔𝑟𝑎𝑝ℎ. 𝑙𝑒𝑛𝑔𝑡ℎ; 𝑖 + +; )

{ 𝑘𝑒𝑦𝑠𝑡𝑟𝑜𝑘𝑒 = 𝑘𝑒𝑦𝑠𝑡𝑟𝑜𝑘𝑒 + 𝑑𝑖𝑔𝑟𝑎𝑝ℎ[𝑖];

𝐾𝑒𝑦𝑠𝑡𝑟𝑜𝑘𝑒𝑇𝑖𝑚𝑒 = 𝑘𝑒𝑦𝑠𝑡𝑟𝑜𝑘𝑒 𝑑𝑖𝑔𝑟𝑎𝑝ℎ. 𝑙𝑒𝑛𝑔ℎ𝑡⁄ ; }

Else

{ “go back to registration page” }

End

3.2.3 Algorithm for User Authentication and Keystroke Verification

User opens the login page.

Begin;

// Test for password and username match

𝒊𝒇 ((𝑢𝑠𝑒𝑟. 𝑝𝑎𝑠𝑠 == 𝑝𝑎𝑠𝑠𝑤𝑜𝑟𝑑)&&(𝑢𝑠𝑒𝑟𝑛𝑎𝑚𝑒 = 𝑢𝑠𝑒𝑟𝑛𝑎𝑚𝑒))

{

//Set threshold

𝒊𝒏𝒕 𝑡ℎ𝑟𝑒𝑠ℎ𝑜𝑙𝑑 = 10; //default threshold for password length is 10

𝒊𝒏𝒕 𝑙𝑒𝑛𝑔𝑡ℎ = 𝑠𝑡𝑟𝑔𝑙𝑒𝑛(𝑢𝑠𝑒𝑟. 𝑝𝑎𝑠𝑠);

𝒊𝒇 𝑙𝑒𝑛𝑔ℎ𝑡 ≥ 11 && 𝑙𝑒𝑛𝑔𝑡ℎ ≤ 20;

{ i𝒏𝒕 𝑡ℎ𝑟𝑒𝑠ℎ𝑜𝑙𝑑 = 15; }

𝒊𝒇 𝑙𝑒𝑛𝑔ℎ𝑡 ≥ 21 && 𝑙𝑒𝑛𝑔𝑡ℎ ≤ 30;

{ i𝒏𝒕 𝑡ℎ𝑟𝑒𝑠ℎ𝑜𝑙𝑑 = 20; }

// Test for keystroke

𝒊𝒇 ((𝑢𝑠𝑒𝑟. 𝑘𝑒𝑦𝑠𝑡𝑟𝑜𝑘𝑒 + 𝑡ℎ𝑟𝑒𝑠ℎ𝑜𝑙𝑑 ≥ 𝑘𝑒𝑦𝑠𝑡𝑟𝑜𝑘𝑒)&&(𝑢𝑠𝑒𝑟. 𝑘𝑒𝑦𝑠𝑡𝑟𝑜𝑘𝑒 − 𝑡ℎ𝑟𝑒𝑠ℎ𝑜𝑙𝑑

≤ 𝑘𝑒𝑦𝑠𝑡𝑟𝑜𝑘𝑒))

{ “Login successful” }

else

{ “Access denied due to keystroke conflict” }


else

{ “Password or username incorrect” }

End

3.2.4 Sales and Stock Processes

This step is after the login, the user click on the process he wants to perform operation on, follow the steps

and submit after wards. The follow lists are the pages on the index page for performing the named operations;

View Transaction

Add Product

Daily Report

Transaction list

Change password

Create User Account

Stock Category

View stock

Branch

Logout


95

3.3 Relational Database Model for the System

Login

User_id int(30) pk

Username Varchar(30) nn

Pass Varchar(30) nn

Name Varchar(30) nn

Role Varchar(30) nn

Branch_id Varchar(30) nn

Keystroke_Time(30) nn

Branch

Branch_id mediumint(40) pk

Name Varchar(40) nn

Branch_manager varchar(40)

Customer

Id int(50) pk

Fullname Varchar(50)

Gender Varchar(30) nn

city Varchar(40) nn

State Varchar(40) nn

Telephone Varchar(40) nn

Address Varchar(50) nn

Category

Cat_id mediumint(9) pk

Category Varchar(40) nn

Products

Product_id mediumint(30) pk

Product_name varchar(50) nn

Unit_price Varchar(30) nn

Category Varchar(40) nn

Quantity Varchar(40) nn

Branch Varchar(40) nn

Barcode Varchar(40) nn

Purchase order

Transact_id mediumint() pk

Date date nn

Customer_name Varchar(50)

nn

Total Varchar(50) nn

Paid Varchar(50) nn

Acc_balance Varchar(50) nn

Invoice Varchar(5000) nn

Operator Varchar(50) nn

Item1 Varchar(100) nn










Branch Varchar(500) nn

Trans_session Varchar(30) nn

Qty Varchar(30) nn

Unit_price Varchar(30) nn

Stock

Id mediumint(50) pk

Item Varchar(40) nn

Unit Varchar(30) nn

Qty Varchar(30) nn

Price Varchar(40) nn

Amt_paid Varchar(40) nn

Balance Varchar(40) nn


Login Table

The user id, username, password, users’ privilege, branch and the keystroke time

are stored on the login table.

4 RESULTS AND DISCUSSION

Create Account

This page is used for the creation of new user account by users with admin level privilege.


97

In this module, the keystroke feature is extracted from the password typed in by the authorized

user when creating account to statistically calculate the Keystroke Time which will be saved in the

database against each user.

Figure 3: The create account page

Branch: the branch selected will be profiled to the new user.

Admin (if it is an admin user)

Creation of Keystroke Dynamics Authentication

This page shows the initial keystroke data of an individual user created which will be stored in the

database after creating the account.


Figure 4: The keystroke created

Figure 5(a): Account creation for user 1


99

Figure 5(b): Keystroke dynamics authentication created for user 1

Figure 6 (a): Account creation for user 2


Figure 6 (b): Keystroke dynamics authentication created for user 2

Login Page

The login page is where users are accessed by identifying and authenticating the user

through the username and password presented by the user. All other pages of the application will

refer the user to the login page for authentication.


101

Figure 7: The application login page

Username: enter username used for registration

Password: is a secret word or alphanumeric used to gain admission into the system

which can only be created by an authorized user.

Error Page for Wrong Login Details

This page shows the error, to inform the user that the username or password doesn’t correspond

with the initially stored in the database. The system only proceeds to keystroke verification only

if the username and password corresponds.


Figure 8: Error page for wrong username or password

Back: it directs the user back to the login page

Error Page for Keystroke Dynamics

This page shows the error, to inform the user about the reason for being unable to login. It will get

the information detail stored in the database against the present one from the login page. If a

malicious transaction is suspected, even if the password is correct, the unauthorized user won’t be

able to login because of the keystroke data saved against each user in the database. The present

keystroke data of an individual user might be more or less than initial keystroke data stored in the

database. The page takes you back to re-login.


103

Figure 9: Error page for keystroke dynamics out of threshold

Back: it directs the user back to the login page

Figure 10(a): User 1 login page


Figure 10(b): User 1 error page for keystroke dynamics authentication.

Figure 11 (a): User 2 login page


105

Figure 11(b): User 2 error page for keystroke dynamics authentication.

Home Page

This page is the initial page of the website, the point of entry to all the information stored within

the application. It contains link to selection of all other available content. This page summarizes

the aim of the work.


Figure 12: Home page

It contains the following options:

Change password: this is where all the staff can change the secret word used, for a

new user and when the user feels it is exposed.

Create account: to create new account for staff.

Manage users: this where the admin user can delete a staff id when resigned or fired

Stock category: is the type of goods available for sales.

View stock: this is where the staff can view the goods available in the stock.

Transaction: is the day-to-day sale.

Add product: is used to add product to the stock.

Daily report: shows daily journal. An admin user can check for all staff transaction

while the sales representative can see his transaction alone.

Barcode Reader: shows transaction using a scanner or manual input of the product

id.


107

Figure 13(a): User 1 login page with correct username, password and keystroke

Figure 13(b): User 1 Home page after Keystroke Dynamics Authentication


Figure 14(a): User 2 login page with correct username, password and keystroke

Figure 14(b): User 2 Home page after Keystroke Dynamics Authentication

Change Password


109

This page allows user to change the password to protect their accounts from fraudulent

transactions. The keystroke data will be extracted from the new password which will be saved

against the user in the database. It is advisable for a user to change their password every three

months to protect their accounts and saves the company from losses.

Figure 15: Change password page

Error Page for Change Password

This page occurs if the new password and the confirm password doesn’t match.


Figure 16: Error page for mis-match password

Logout: exit the application.

Back: it directs the user back to the change password page


111

5 CONCLUSION

In this work, a behavioral biometrics, keystroke dynamics authentication solution is proposed on

the usual password authentication used for authorization. The inventory of different steps in

business enterprises where taken with emphasis on stocking, transaction and staffing processing

to develop an e-commerce system that a multifactor authentication is implemented on with the aim

of enhancing the security system. The researchers implemented a behavioral biometric method,

keystroke dynamics authentication static method on the usual passwords authentication for login

to produce a multifactor authentication system that helps tackle the variety of security frauds

involved in user login process, whether in house or from the outside by using the typing style of

the authorized users which has been proven unique with the password. The researchers suggested

a more secured method for Authentication, as well as a business solution for SMEs. This work

can also detect whether a user is authorized or fraudulent which makes it secure. Experimental

results show that the performance and effectiveness of this system demonstrate the usefulness of

implementing it. The system is also scalable for handling large volumes of transactions. This new

system meets user needs, easy to understand and can be used by both educated (professionals) and

uneducated (local traders) users. This system has qualities such as reduction in the cost of adding

new features to the existing system since the keyboard is the only hardware required and also built

in maintenance view with low execution and response time.


REFERENCES

ABERNETHY, M., AND RAI, S. 2012. Applying Feature Selection to Reduce Variability in

Keystroke Dynamics Data for Authentication Systems. In proceedings of the 13th Australian

Information Warfare and Security Conference, Perith, Western Australia, December, 2012, 17-

23.

ABUALGASIM, S.D., AND OSMAN, I. 2011. An Application of the Keystroke Dynamics

Biometric for Securing PINs and Passwords. World of Computer Science and Information Technology

Journal (WCSIT) ISSN: 2221-0741, 1(9), 398-404.

AL-JARRAH, M.M. 2012. An Anomaly Detector for Keystroke Dynamics Based on Medians

Vector Proximity. Journal of Emerging Trends in Computing and Information Sciences, 3(6), 988-

993.

VASTHI, S., AND SANWAL, T. 2016. Biometric Authentication Techniques: A Study on

Keystroke Dynamics. International Journal of Scientific Engineering and Applied Science

(IJSEAS) ISSN: 2395-3470, 2(1), 215-221.

BAJAJ, S., AND KAUR, S. 2013. Typing Speed Analysis of Human for Password Protection

(Based On Keystrokes Dynamics). International Journal of Innovative Technology and Exploring

Engineering (IJITEE) ISSN: 2278-3075, 3(2), 88-91.

DAVENPORT, T. 1998. Putting the Enterprise into the Enterprise System. Harvard Business

Review, 76(4), 121-131.

EPP, C. 2010. Identifying Emotional States through Keystroke Dynamics. A thesis submitted to

the college of Graduate Studies and Research in partial fulfillment of the Requirements for the

Degree of Master of Science in the Department of Computer Science, University of Saskatchewan

Saskatoon, Canada.

FERREIRA, J., SANTOS, H., AND PATRAO, B. 2011. Intrusion Detection through Keystroke

Dynamics. In proceedings of the 10th European Conference on Information Warfare and Security,

Tallinn, Estonia, July 2011, 81-90.


113

GIOT, R., EL-ABED, M., AND ROSENBERGER, C. 2014. Keystroke Dynamics Authentication.

Biometrics, In Tech, Chapitre 8. 2011. 978-953-307-618-8. <10.5772/170647>. <hal-00990373>

157-182.

GIOT, R., EL-ABED, M., AND ROSENBERGER, C. 2009. Keystroke Dynamics Authentication

for Collaborative System. The IEEE International Symposium on collaboration Technologies and

Systems (CTS), Baltimore, United States, May 2009, IEEE Computer Society, 172-179.

<10.1109/CTS2009.5067478> .<hal-00432764>

KARNAN, M., AND KRISHNARAJ, N. 2012. A Model to Secure Mobile Devices Using

Keystroke Dynamics through Soft Computing Techniques. International Journal of Soft

Computing and Engineering (IJSCE) ISSN: 2231-2307, 2(3).

MASOCHA, R., CHILIYA, N., AND ZINDIYE, S. 2011. E-banking adoption by customers in the

rural milieus of South Africa: A case of Alice, Eastern Cape, South Africa. African Journal of

Business Management, 5(5), 1857-1863.

MODI, J., UPADHYAY, H.G., AND THAKOR, M. 2014. Password less Authentication Using

Keystroke Dynamics: A Survey. International Journal of Innovative Research in computer and

Communication Engineering, 2(11), 7060-7064.

MONROSE, F., AND RUBIN, A.D. 1999. Keystroke Dynamics as a Biometric for Authentication.

Future Generation Computer Systems, 16(2000), 351-359.

SAWANT, M.M., NAGARGOJE, Y., BORA, D., SHELKE, S., AND BORATE, V. 2013.

Keystroke Dynamics: Review Paper. International Journal of Advanced Research in Computer and

Communication Engineering, 2(10), 4018- 4020.

SCOTT, J. E., AND VESSEY, I. 2002. Managing Risks in Enterprise Systems Implementations.

Communication of the ACM, 45(4), 74-81.

SHUKLA, P., AND SOLANKI, R. 2013. Web Based Keystroke Dynamics Application for

Identifying Emotional State. International Journal of Advanced Research in Computer

Communication Engineering, 2(11), 4489-4493.

SINGH, S., AND ARYA, K.V. 2011. Key Classification: A New Approach in Free Text Keystroke

Authentication System. In proceedings of the 3rd Pacific- Asia Conference on Circuits and

Communication Systems (PACCS), Wuhan, China, July, 2011, 1-5.


VANDOMMELE, T. 2010. Biometric authentication today. Seminar on Network Security.

Retrieved from: http://www.cse.hut.fi/en/publications/B/11/papers/vandommele.pdf

http://www.cse.hut.fi/en/publications/B/11/papers/vandommele.pdf

international journal of computing and ict...

Documents