ieee internet of things journal, vol. aa, no. bb, mmmm...

IEEE INTERNET OF THINGS JOURNAL, VOL. AA, NO. BB, MMMM 2017 1

Mobile Edge Computing: A SurveyNasir Abbas, Yan Zhang, Senior Member, IEEE, Amir Taherkordi, Member, IEEE, and Tor Skeie Member, IEEE

Abstract—Mobile Edge Computing (MEC) is an emergentarchitecture where cloud computing services are extended to theedge of networks leveraging mobile base stations. As a promisingedge technology, it can be applied to mobile, wireless and wirelinescenarios, using software and hardware platforms, located atthe network edge in the vicinity of end-users. MEC providesseamless integration of multiple application service providersand vendors towards mobile subscribers, enterprises and othervertical segments. It is an important component in the 5Garchitecture which supports variety of innovative applicationsand services where ultra low latency is required. This paper isaimed to present a comprehensive survey of relevant research andtechnological developments in the area of MEC. It provides thedefinition of MEC, its advantages, architectures, and applicationareas; where we in particular highlight related research andfuture directions. Finally, security and privacy issues and relatedexisting solutions are also discussed.

Index Terms—Mobile Edge Computing, Mobile Cloud Com-puting, Fog Computing, IoT.

I. INTRODUCTION

THE prevalence of mobile terminals, such as smartphonesor tablet computers, has an uttermost effect on mo-

bile and wireless networks, triggering challenges for mobilenetworks worldwide [1] [2]. This class of networks has toendure low storage capacity, high energy consumption, lowbandwidth, and high latency [3]. Moreover, exponential growthof the emerging Internet-of-Things (IoT) technology is fore-seen to further stumble cellular and wireless networks [4].Mobile Cloud Computing (MCC), as an integration of cloudcomputing and mobile computing, has provided considerablecapabilities to mobile devices and empowered them withstorage, computation and energy resources offered by thecentralized cloud [5] [6]. However, popping up a myriad ofmobile devices, MCC is encountering noticeable challenges,such as high latency, security vulnerability, low coverage andlagged data transmission. These challenges can become moredifficult to address in the case of next generation mobilenetworks (e.g., 5G) [7]. Moreover, MCC is not suitable forscenarios involving real-time applications and guaranteeinghigh Quality of Service (QoS). According to the recent reportpresented by Cisco Visual Networking Index, 11.6 billionmobile-connected devices will be used by 2020 [2]. The trendof increase in mobile devices usage is fundamentally drivenby the augmentation of mobile users and mobile applicationsdevelopment (e.g., iPhone apps and Google apps) [8] [9].

N. Abbas was with the Department of Informatics, University of Oslo,Norway email: ([email protected])

Y. Zhang (corresponding author) and A. Taherkordi are with the Departmentof Informatics, University of Oslo, Norway email: ([email protected],[email protected])

T. Skeie is with Simula Research Laboratory and the Department ofInformatics, University of Oslo, Norway email: ([email protected])

Manuscript received November 30, 2016; revised March 13, 2017.

In the era of computing paradigms, edge computing (alsoknown as fog computing) [10], has begun to be of paramountsignificance, especially Mobile Edge Computing (MEC) inmobile cellular networks. The main purpose of mobile edgecomputing is to address the challenges that are hailed fromMCC systems. MEC empowers MCC by deploying cloudresources, e.g., storage and processing capacity, to the edgewithin the radio access network. This provides the end-user with swift and powerful computing, energy efficiency,storage capacity, mobility, location, and context awarenesssupport [11] [12]. Previously, the technology at the edge ofthe Internet known as cloudlet technology has been introducedto deploy mobile cloud services, however it was inadequatebecause of its limited WiFi coverage. In a highly compu-tational environment, cloudlets can efficiently process thecomputationally-intensive tasks offloaded from devices [12].Alternatively, MEC is equipped with better offloading tech-niques that characterize the network with low-latency andhigh-bandwidth.

Although MEC technologies and reported research con-tributions are still young and limited, providing a thoroughoverview of the state-of-the-art in MEC development willprovide useful insights to the current status of this area andhelp to uncover potential research directions. This basicallyshapes the contribution of this paper, which is surveying MEC.There exist a few MEC survey reports in the literature suchas [13] and [14]. However, the former mainly analyzes securitythreats and challenges that affect different edge paradigms,such as fog computing, mobile edge computing, and mobilecloud computing. The latter provides a brief overview ofdifferent attributes of MEC and identifies the major openresearch challenges in MEC, while the detailed and thoroughstudy of different design aspects and research trends is largelymissing in this survey work.

This paper presents an extensive survey on mobile edgecomputing focusing on its general overview. Section 2 givesan overview of MEC encompassing definition, concepts, ar-chitectures and its advantages. Sections 3 and 4 list MECapplications and emerging scenarios. Section 5 presents state-of-the-art research on MEC with respect to computationaloffloading, latency, storage, and energy efficiency. Then, re-search infrastructures are presented in Section 6 which hasbeen referred to as MEC testbeds in recent studies. Sections7 and 8 outline security and privacy issues, and securitymechanism. Finally Section 9 discusses MEC open issues.The list of acronyms and abbreviations used in this paper aresummarized in Table 1.

II. MOBILE EDGE COMPUTING OVERVIEWThe term mobile edge computing was standardized by

European Telecommunications Standards Institute (ETSI) and


TABLE IACRONYMS/ABBREVIATIONS

1G First Generation2G Second Generation3G Third Generation4G Fourth Generation5G Fifth Generation5GTN 5th Generation Test NetworkAPI Application Program InterfaceAR Augmented RealityASP Application Service ProviderBS Base StationBSC base station controllernDCs Nano Data CentersDNS Domain Name ServerDOS Denial of ServiceEAB Edge Accelerated Web BrowsingEEG ElectroencephalogrameNodeB Evolved Node BETSI European Telecommunications Standards InstituteFC Fog ComputingFN Fog NodeGPRS General Packet Radio ServiceGPS Global Positioning SystemIaaS infrastructure-as-a-ServiceIDS Intrusion Detection SystemIoT Internet of ThingsISG Industry Specification GroupLBS Location Based ServiceLTE Long Term EvolutionM2M Machine-to-machineMCC Mobile Cloud ComputingMNO Mobile Network OperatorMU Mobile UsersNFC Near field communicationNILL Non-Intrusive Load LevelingNP-hardness non-deterministic polynomial-time hardOTT Over-the-topQoE Quality of ExperienceRAN Radio Access NetworkSRAN Service-Aware RANSAE System Architecture EvolutionSCADA Supervisory Control Data AcquisitionSDCompute Software Defined ComputeSDN Software Defined NetworkSDSec Software Defined SecuritySDStorage Software Defined StorageSRAN Service-Aware RANTDMA Time Division Multiple AccessUE User EquipmentUMTS Universal Mobile Telecommunication SystemUTRAN UMTS radio access networkVDTNs Vehicular Delay-Tolerant NetworksWAN Wide Area NetworkWSAN Wireless Sensor Actuator Network

Industry Specification Group (ISG). The ISG group includesNokia Networks, Intel, Vodafone, IBM, Huawei and NTTDOCOMO, to name a few. MEC is also acknowledged byEuropean 5G PPP (5G Infrastructure Public Private Partner-ship) as a prime emerging technology for 5G networks [15].

A. Definition of Mobile Edge Computing

According to ETSI, mobile edge computing is definedas [15]:

“Mobile Edge Computing provides an IT service environ-ment and cloud computing capabilities at the edge of themobile network, within the Radio Access Network (RAN) andin close proximity to mobile subscribers.”

Mobile edge computing offers cloud computing capabilitieswithin the radio access network. Allowing direct mobile trafficbetween the core network and the end-user, instead, MECconnects the user directly to the nearest cloud service-enablededge network. Deploying MEC at the base station enhancescomputation and avoids bottlenecks and system failure [16],[7].

According to the white paper published by ETSI, mobileedge computing can be characterized by [17]:

1) On-Premises: MEC platforms can run isolated fromthe rest of the network, while they have access tolocal resources. This is very important for machine-to-machine scenarios. The MEC property of segregationfrom other networks also makes it less vulnerable.

2) Proximity: Being deployed at the nearest location, mo-bile edge computing has an advantage to analyze andmaterialize big data. It is also beneficial for compute-hungry devices, such as augmented reality, video ana-lytics, etc.

3) Lower latency: Mobile edge computing services aredeployed at the nearest location to user devices, isolatingnetwork data movement from the core network. Hence,user experience is accounted high quality with ultra-lowlatency and high bandwidth.

4) Location awareness: Edge-distributed devices utilizelow-level signaling for information sharing. MEC re-ceives information from edge devices within the localaccess network to discover the location of devices.

5) Network context information: Applications providingnetwork information and services of real-time networkdata can benefit businesses and events by implementingMEC in their business model. On the basis of RANreal-time information, these applications can estimatethe congestion of the radio cell and network bandwidth.This will help them in future to make smart decisionsfor better delivery of services to customers.

B. Related Concepts and Technologies

There are some terms similar to mobile edge computing,such as mobile cloud computing, local cloud, cloudlets andfog computing [18]. In the following, we carefully discussthese terms.

• Mobile Cloud Computing (MCC) generally integrates allthe advantages of mobile computing, cloud computingand mobile internet [19]. The main focus of cloudcomputing is to enable isolated virtualized computing,storage and communication resources that are leveragedby end-users [20]. Some examples of cloud computinginfrastructures and platforms are Amazon EC2, MicrosoftAzure, Google, and Aneka. Mobile cloud computingenables resources on demand, such as network, server,application, storage and computing resources in a mobileenvironment [21]. MCC also focuses on resource man-agement that could be easily manageable [20]. In a MCCinfrastructure, the centralised cloud servers are locatedfar off from end devices, therefore are less productive incomputation intense environments. For example, mobile


applications connected to the cloud may face networklatency or disconnections while mobile applications areused.

• Local Cloud is administered by internal or externalsources explicitly intended for a group or institution [22].Local cloud is deployed in a local network that coordi-nates with its remote cloud server to promote data privacy.It is enabled by installing software on the local serverthat is integrated with the cloud server. However, localcloud is favorable in terms of communication delay butit is subject to some computational limitations due to itssparse resources [23].

• Cloudlet is a small-box data center that is normallydeployed at one wireless hop away from mobile devices,such as public places like hospital, shopping center,and office building to facilitate a convenient approachas shown in Figure 1 [24]. Several units of multi-core computers form a cloudlet that is connected toremotely located cloud servers. Cloudlet is brought asa promising solution to overcome the distant Wide AreaNetwork (WAN) latency and cellular energy consump-tion problem by utilizing cellular data connectivity tonear-located cloud servers [25]. The primary focus ofcloudlet is to bring cloud technologies closer to the end-user, providing support to resource- and latency-sensitiveapplications [26]. Cloudlets utilize technologies, such asWiFi located at one hop or multiple hops away at theedge of the Internet. Therefore, it is dependent on robustand uninterrupted Internet connection. Herein, there aresome security and privacy issues that involve the accessprivacy services, such as e-commerce websites [27].

Fig. 1. Cloudlet

• Fog Computing is also known as edge computing, sup-porting ubiquitous connected devices. The Fog computingterm was created by CISCO systems to bring cloudservices to the edge of an enterprise network. In Fog

computing, processing is mainly carried out in the localarea network and in an IoT gateway or a fog node. Fogcomputing offers the benefit to allow single processingdevices to gather data from different sensors and actaccordingly. For example, a smart robotic vacuum cleanerreceives data from multiple sensors installed in a housethat are capable to detect any dirt and send appropriatecommands to the vacuum cleaner to react accordingly.Fog computing offers much low latency as compared tocloud computing, located far from the end-user. However,fog computing has some limitations due to its dependencyon wireless connection which has to be live in order toperform complex actions. Fog computing and MEC termsare widely used interchangeably but they differ in someways. For example, in a fog computing environment,intelligence is at the local area network level whichis processed at the fog node or the IoT gateway [28].Therefore, there is a rising trend in wireless networksfor IoT and machine-to-machine (M2M) communication.However, in mobile edge computing environments, intel-ligence, communication capability and processing powerare pushed to the RAN, thereby MEC is becoming morepopular in 4G and future 5G networks.

C. Architectures of Mobile Edge Computing

Mobile edge computing functions are mostly found withinthe Radio Access Network (RAN), thereby prior to discussingMEC architectures, we need to first understand the history andgeneral architecture of cellular network communication froma RAN perspective.

1) History and Role of RAN in Cellular Networks: Backin early 1980s, the first commercial cellular network (1Ggeneration) was introduced with the compliance of analogmodulation and mobility support, which later was replaced by2G because of its digital radio signaling capability using TimeDivision Multiple Access (TDMA). 2G networks were knownfor better voice quality which was achieved by leveraging dig-ital technology for better voice quality. Later, 3G was releasedwith better data transfer rate and multimedia applicationscoherence using RAN with limited data support [29]. Withan accustomed support of mobile internet using RAN Long-Term Evolution (LTE), 4G got an edge over other wirelessmobile telecommunication technologies providing the bestuser experience [30].

RAN is part of the cellular network communication sys-tem infrastructure, facilitating the connection between mobilephones or any wireless controlled machines with the mo-bile core network [31]. In traditional cellular radio systems,wireless user equipment connect through RAN to the mobileoperator networks. User equipment include mobile stations,laptops, etc. RAN covers the wide geographical area dividedinto several cells and each cell is integrated with its basestation. Base stations are typically connected to each othervia microwave or landlines to the Radio Network Controller(RNC), which is also known as the Base Station Controller(BSC). RNC is responsible to control the base station nodeand also carry out some mobile management functions. Most


of the encryption is performed before sending user data tothe core network. The RNCs are connected with one or twoback haul networks. Cellular networks have become moreefficient than before because the LTE technology provideshigh-speed wireless communication radio access networkswith low-latency and high-bandwidth. System ArchitectureEvolution (SAE) of RAN LTE core conforms heterogeneousnetworks and legacy systems, such as air interfaces of GeneralPacket Radio Service (GPRS) or Universal Mobile Telecom-munications (UMTS) [9]. The UMTS is a third generationsystem that may depend on the global system for mobilecommunication (GSM) developed in Europe.

The generic view of cellular networks is depicted in Figure2, where the core network is wire-connected (e.g., IP/Ethernet)with RAN and RAN is wireless-connected to user devices.RAN connects base station with backhaul network through theEthernet interface which supports a high data transfer rate [32].

Fig. 2. Cellular Architecture

In the past, IP has grown from the Internet, to organizationnetworks and increasingly adopted by the LTE network. The IPtraffic between RAN and core is encapsulated with the GPRStunneling protocol with an IPsec encryption [33]. This hasprohibited IT services to be inserted at the nearest locationto the end-users. Moreover, mobile operators are reluctant todeploy applications due to the risk of denial of mobile servicesor performance decrease.

Recently, the concept of Cloud Radio Access Network (C-RAN) has been proposed by a few operators. C-RAN promisesa centralized processing, collaborative radio, real-time cloudcomputing, and power efficient infrastructure [34]. In particu-lar, it aggregates all base station computational resources intoa central pool. In C-RAN, the radio frequency signals fromgeographically distributed antennas are collected by remoteradio heads and transmitted to the cloud through an opticaltransmission network. Using C-RAN, the number of cell siteswill be reduced and the user will be offered better services,while it maintains similar coverage and reduces operatingexpenses.

2) Three-Layer Architecture: MEC is a layer that residesbetween the cloud and mobile devices. Therefore, the in-frastructure is derived as a three-layer hierarchy; the cloud,MEC and mobile devices [35]. Mobile edge computing mostlycomplies with cloud computing to support and enhance per-formance of the end devices. The formation of a three-layerservice model is depicted in Figure 3.

The general architecture of mobile edge computing is de-picted in Figure 4. As shown, different types of mobile devices

Fig. 3. Three-layer architecture

and sensors in, e.g., IoT, big data and social platforms, areconnected to the core network (i.e., mobile Internet) throughthe edge network, i.e., the radio access network and MEC, andthe core network is connected to the private cloud network.With the evolution of LTE-based RAN, it has become morefeasible to deploy MEC which brings cloud services near to themobile subscribers. Therefore, as shown in the architecturalmodel in Figure 4, each edge platform represents an edgecloud with applications and services specific to the targetmobile environment.

MEC constitutes geo-distributed servers or virtual serverswith built-in IT services. These servers are implementedlocally at mobile user premises, e.g., parks, bus terminals,and shopping centers [35]. MEC may utilize cellular networkelements, such as the base station, WiFi access point, or femtoaccess point (i.e., low power cellular base station). MEC maybe deployed at a fixed location, for example, in a shoppingcenter or on a mobile device located in any moving object,e.g., car or bus. MEC can be deployed at a LTE base station(eNodeB) or a multi-technology (3G/LTE) cell aggregationsite. The multi-technology cell aggregation site can be locatedboth indoor or outdoor. To push intelligence at the basestation and to effectively optimise RAN services, mobile edgecomputing technology develops an energetic ecosystem and anew value chain that allows intelligent and smart services atnearby locations to the mobile subscribers.

To summarize, the key value proposition of MEC is thatit offers cloud computing by pushing cloud resources, suchas compute, network and storage to the edge of the mobilenetwork in order to fulfil application requirements that arecompute hungry (e.g., games applications), latency-sensitive(e.g., augmented reality applications) and high-bandwidth de-manding (e.g., mobile big data analytics).

D. Advantages of Mobile Edge ComputingAs already discussed in previous sections, there are several

benefits associated with mobile edge computing which areturning out to be promising for both Mobile Network Op-erators (MNOs) and Application Service Provider (ASP). In


Mobile Internet

edge

edge

edge Enterprise Network

Private Cloud

DMZ

API

Apps Service

Apps Service

Apps Service

Developers & Customers

Big Data

IOT Sensors

Social and Internet Data Sources

Fig. 4. Mobile edge computing architecture

addition, they are beneficial to content providers, Over-The-Top (OTT) players, network equipment vendors, and middle-ware providers [36], [9]. MEC concept focuses on importantmetrics, such as delay and high-bandwidth that are accom-plished by limiting data movement to MEC servers than tocentralized servers with severe latency cost. Moreover, powerconsumption is also one of the main concerns. Computationaltasks are migrated to external resource-rich systems to increasethe battery lifetime of user equipment. In addition, distributedvirtual servers provision scalability and reliability.

With respect to the actors (MNOs, ASPs and end-users),MEC benefits include [37], [9]:

• Mobile network operators could enable RAN access tothird party vendors to deploy their applications and ser-vices in a more flexible and agile manner. These enablingservices could generate revenue by charging based on theservices used, such as storage, bandwidth, and other ITresources. OTT services and DVR services offered bycable operators may likely be faster since their servicescould reside in MEC servers.

• Application service providers could gain profit by aMEC-enabled Infrastructure-as-a-Service (IaaS) platformat the network edge which makes ASPs services scalablealong with high bandwidth and low latency. ASPs couldalso get real-time access to the radio activity that maydevelop more capable applications. RAN is revamped intoService-Aware RAN (SRAN) which provides the locationinformation of the subscriber, cell load, and networkcongestion.

• End-users could experience fast computational applica-tions through offloading techniques that are handled byMEC servers within RAN. In addition, tight RAN assim-ilation and physical close servers could improve the userQuality of Experience (QoE), such as high throughputbrowsing, video caching, better DNS, etc.

III. APPLICATIONS

Although the MEC architecture is a new revenue stream formobile operators that has not matured sufficiently, we witnessa few application areas adopting Edge Computing (e.g., FogComputing) as it has been focused by recent articles [15], [38].Some recognized applications include Augmented Reality andContent Delivery.

A. Augmented Reality (AR)

In the era of mobile technology, augmented reality applica-tions have recently adopted mobile technology, such as Layar,Junaio, Google Goggles, and Wikitude [39]. AR enables realenvironment user experience by combining real and virtualobjects existing simultaneously [40], [41]. Recent AR appli-cations have become adaptive in sound and visual components,such as news, TV programs, sports, object recognition, games,etc [42]. However, AR systems usually demand high comput-ing power for task offloading, low latency for better QoE, andhigh bandwidth that is conducive to sustaining interminableIT services.

Edge computing infrastructures have been recognized tobe a niche for latency-sensitive applications in the AR do-main [43]. They empower AR systems, for example by max-imizing throughput by pushing intelligence to the edge of thenetwork instead of relying on the core network. Therefore,offloading computation-intensive tasks at the nearest cloudletis more optimized and efficient, enhancing user experience.

One example of AR applications is Brain Computer Interac-tion that works by detecting human brainwaves [38]. The datais received by EEG Bio-sensors in real-time acquiring largecomputational tasks handled by MEC and cloud computingplatforms.

B. Content Delivery and Caching

The edge computing technology plays a key role in website performance optimization, such as caching HTML content,reorganizing web layout and resizing web components. Theuser makes HTTP requests that pass through the edge server.This server handles user requests by performing number oftasks to load web pages on the user device interface. Theserequests and responses are time efficient as the edge serveris deployed close to the edge devices. The edge computinginfrastructure is time efficient as compared to the traditionalInternet infrastructures where user requests are handled by theservers that are distantly placed at the service provider side. Inaddition, edge computing also analyses network performanceduring on and off peak hours. For example, under congestednetwork conditions where several users are streaming video atthe same time, the graphics resolution is decreased to minimalto accommodate every user averting any denial of service orjitter.

IV. EMERGING APPLICATION SCENARIOS

It is important to stay ahead of the curve to apprehendmobile technology trends. In this section, emerging applicationscenarios of MEC are presented which are recently discussedin the ETSI white paper [15], such as video analytics andmobile big data. Several research papers [18], [44], [10], [45]have referred to MEC scenarios in connected vehicles, smartgrid and Wireless Sensor and Actuator Networks (WSAN).Furthermore, in [46], the application area is expanded to smartbuilding control and Software-Defined Network (SDN), aswell as to ocean monitoring [47].


A. Healthcare

Science and technology in health domain is a substantialresearch area for many researchers [48]. Like other industries,healthcare can also be aided by edge computing, e.g., patientssuffering from strokes fall. According to the stroke statistics,someone in the U.S. has a stroke about once every 40 sec-onds [49]. Falls are common among stroke patients who suffermostly from hypoglycemia, hypotension, muscle weakness,etc. According to recent research, one third of the strokes couldpossibly be averted by early mitigating the fall incidents [50].In order to detect and prevent fall, a large body of researchhas been carried out, for example, by introducing humancomputer interaction devices, such as smartphone, smart watchand Google glass, but certain limitations still exist.

Recently, researchers have proposed a smart healthcareinfrastructure called U-Fall, that exploits smartphones byengaging edge computing technology. U-Fall is based on afall detection algorithm that is designed using accelerationmagnitude values and non-linear time series analysis [48],[38]. U-fall senses motion detection with the help of smartdevice sensors, such as gyroscopes and accelerometers. U-Fallintelligently maintains integrity between the smartphone andthe cloud server to ensure real-time detection. In addition, theproposed infrastructure is capable to deliver accurate resultswhich make it more reliable and dependable.

Furthermore, the three-tier architecture that includes rolemodel, layered-cloud architecture, and mobile edge computingcan help health advisers to assist their patients, independentof their geographical location. MEC enables smartphones tocollect patient physiological information, e.g., pulse rate, bodytemperature, etc., from smart sensors and send it to the cloudserver for storage, data sync and sharing. Health advisershaving access to the cloud server can immediately diagnosepatients and assist them accordingly [51].

B. Video Analytics

Surveillance cameras in old times were used to stream databack to the main server and then the server decided how toperform data management. Due to the growing ubiquity ofsurveillance cameras, the traditional client-server architecturemight not be able to stream video coming from million of de-vices and therefore, it will stress the network. In this scenario,MEC will be beneficial by implementing intelligence at thedevice itself which is programmed to send data to the network,when there is a motion detection. In addition, MEC-enabledsurveillance cameras can be beneficial for several applications,such as traffic management applications which can detecttraffic jam or an accident on the basis of traffic patterns.The application can also be helpful for face recognition, forexample, if someone commits a crime then his photo can betransferred to these intelligent cameras to trace the culprit [52],[53], [52].

C. Mobile Big Data Analytics

Mobile phone technology is valued a growth-engine forsmall, medium and large enterprises, and also has widespread

social connotation. The ubiquity of mobile phones and theirbig data coming from applications and sensors, such as GPS,accelerometer, gyroscope, microphone, camera and bluetoothare stressing the network bandwidth [54]. Big data consistsof large and complex data sets that is generated by dataprocessing applications, sensors, devices, video and audiochannels, and web and social media [55], [56]. These datasets may be structured or non-structured and may not bepossible to process them by a single machine [57]. Big datais of paramount importance to businesses because it extractsanalytics and useful information that may benefit to differentbusiness segments [58]. Big data analytics is a process ofextracting meaningful information from raw data which couldbe helpful for marketing and targeted advertising, customer re-lations, business intelligence, context-aware computing, healthcare, etc. [59], [60].

Implementing MEC near the mobile devices can elevate bigdata analytics with the help of network high bandwidth andlow latency. For example, instead of using the typical pathfrom an edge device to the core network, big data can becollected and analyzed at the nearest MEC environment. Theresult of big data analytics can then be passed to the corenetwork for further processing. This scenario will perhaps alsoaccommodate data coming from several IoT devices for bigdata analytics [61].

D. Connected Vehicles

Vehicles are facilitated with an internet access that al-lows them to connect with other vehicles on the road. Theconnection scenario can either be vehicle-to-vehicle, vehicleto access point or access point to access point. DeployingMEC environments along side the road can enable two-waycommunication between the moving vehicles. One vehicle cancommunicate with the other approaching vehicles and informthem with any expected risk or traffic jam, and the presence ofany pedestrians and bikers. In addition, MEC enables scalable,reliable and distributed environments that are synced with thelocal sensors [62].

E. Smart Grid

A smart grid infrastructure is an electrical grid that consistsof several components, such as smart appliances, renewableenergy resources, and energy efficiency resources. Smart me-ters that are distributed over the network are used to receiveand transmit measurements of the energy consumption [63].All the information collected by the smart meter is supervisedin Supervisory Control And Data Acquisition (SCADA) sys-tems that maintain and stabilize the power grid. Moreover,distributed smart meters and micro grids, integrated withMEC, can support SCADA systems. For example, in thisscenario, MEC will balance and scale the load according tothe information shared by other micro grids and smart meters.

F. Wireless Sensor and Actuator Networks

Wireless Sensors and Actuator Networks (WSAN) are sen-sors that are used for surveillance, tracking, and monitoring


of physical or environment situations, e.g., light intensity,air pressure, and temperature [64]. MEC-enabled actuatorsautonomously manage measurement process by developing anactive feedback loop system. For example, air vent sensorsmanage air pressure flowing in and out of the mine to saveminers from any emergency. These sensors consume very lessenergy and bandwidth with the help of MEC.

G. Smart Building Control

Smart building control systems consist of wireless sensorsthat are deployed in different parts of buildings. Sensorsare responsible for monitoring and controlling building en-vironments, such as temperature, gas level or humidity. Ina smart building environment, sensors installed with MECbecome capable of sharing information and become reactive toany abnormal situation. These sensors can maintain buildingatmosphere on the basis of collective information receivedfrom other wireless nodes. For example, if humidity is detectedin the building, MEC can react and perform actions to increaseair in the building and blow out the moisture.

H. Software-Defined Networks

Software-Defined Networks (SDN) are an innovation tocomputer networking that separates control layer and the datalayer [65]. Data layer contains user-generated messages andis responsible to forward them using the forwarding tablesprepared by the control layer [66]. This is managed by acentralized control system. The MEC concept along with SDNcan make centralized control more efficient and reliable, e.g.,in vehicle-to-vehicle connectivity the ratio of packet loss canbe resolved.

I. Ocean Monitoring

Scientists are researching to cope with any ocean cata-clysmic incidents and know the climate changes in advance.This can help to react quickly and mitigate to prevent anydisastrous situation. Sensors deployed at some locations inthe ocean transmit data in great quantity which requires largecomputational resources [47]. The data handled by the cloudmay introduce delays in the transmission of live forecast. Inthis scenario, MEC can play a vital role to prevent any dataloss or delay in sensor data transmission.

V. STATE-OF-THE-ART RELATED RESEARCH

In this section, we present several research efforts carriedout in the area of MEC recently.

A. Computational Offloading

In computer science, computation offloading is the processof migrating computing tasks to external sources, such asclouds, grids or clusters [67]. Computation offloading is a solu-tion to enhance the capacity of mobile devices by transferringcomputation to higher resourceful servers that are located at adifferent location [68]. The emergence of resource-demandingapplications, such as 3D games will continue to demand

more mobile resources. Improvement of mobile devices andnetworks will still not be able to cope up with the trend indemand. Therefore, mobile devices will always have to com-promise with their limited resources, such as resource-poorhardware, insecure connections and energy-driven computingtasks [69]. For example, editing video clips on a mobile phonerequires a large amount of energy and computation which isprovided with some limitations as compared to desktop orlaptop. To deal with these constraints, many researchers havestudied computation offloading to resource-rich platforms,such as the cloud [70], [71], [72].

In 2015, Takahashi et al. [73] proposed Edge Acceleratedweb Browsing (EAB) prototype designed for web applicationexecution using a better offloading technique. The purpose ofEAB is to improve user experience by pushing applicationoffloading to the edge server which is implemented within theRAN. EAB-frontend at the client-side retrieves the renderedweb content which is processed in the EAB server, whereas,audio and video streams travel through the EAB-backendand are decoded depending on the capabilities of the clienthardware.

In 2016, Chen et al. [74] designed an efficient computa-tion offloading model using a game theoretic approach ina distributed manner. Game theory is a persuasive tool thathelps simultaneously connected users to make the correctdecision when connecting a wireless channel based on thestrategic interactions. If all user devices offload computationactivities using the same wireless channel, it might causesignal interference with each other and wireless quality re-duction. Specifically, the game theory targets the NP-hardproblem of computation offloading incurred by multi-usercomputation offloading and provides a solution by attainingNash equilibrium of multi-user computation offloading game.

In 2015, Sardellitti et al. [75] proposed an algorithm-baseddesign, called Successive Convex Approximation (SCA). Thisalgorithm optimizes computational offloading across denselydeployed multiple radio access points. The authors consideredthe MIMO multicell communication system where severalmobile users request for their computational tasks to becarried at the central cloud server. They first tested a singleuser offloading computational task on the cloud server whichresulted in the non-convex optimization problem. In the multi-user scenario, the SCA-based algorithm attained local optimalsolution of the original non-convex problem. According tothe formulation results, authors claimed their algorithms tobe surpassed disjoint optimization schemes. Moreover, theystated that the proposed SCA design is more suitable forapplications acquiring high computational tasks and minimizesenergy consumption.

In 2016, Zhang et al. [76] proposed the contract-basedcomputation resource allocation scheme. This scheme im-proves the utility of vehicular terminals which intelligentlyutilize services offered by MEC service providers under lowcomputational conditions. The MEC provider receives the pay-ment from vehicles on the basis of the computation task theyoffloaded to the MEC servers. Using a wireless communicationservice, information of the contract and payment informationis broadcasted to the vehicles on the road.


In 2015, Habak et al. [77] proposed the FemtoCloud systemwhich forms a cloud of orchestrated co-located mobile devicesthat are self-configurable into a correlative mobile cloudsystem. A FemtoCloud client computing service is installed oneach mobile device to calculate device computing capabilityand capacity for sharing with other mobile devices, and energyinformation. Mobile properties are built and maintained insidea user profile that is shared in a mobile cluster connectedto a cloudlet or a control device that is available in a WiFinetwork. Intensive computational tasks in the form of codesare sent to cloudlets to leverage the computational capacityof other connected mobile devices. The FemtoCloud model isdesigned to reduce the computational load from the centralisedlocation and bring it to the edge of the mobile network.

B. Low Latency

MEC is one of the promising edge technologies that im-proves user experience by providing high bandwidth and lowlatency.

In 2016, Abdelwahab et al. [78] proposed REPLISOMwhich is a edge cloud architecture and LTE enhanced memoryreplication protocol to avoid latency issues. LTE bottleneckoccurs due to allocating memory to a large number of IoTdevices in the backend cloud servers. These devices of-fload computational tasks by replicating and transmitting tinymemory objects to a central cloud which makes IoT to bescalable and elastic. The LTE-integrated edge cloud providesits compute and storage resources at the edge to resource-intensive services. Thus, the proposed REPLISOM reduces thestress of LTE by intelligently scheduling memory replicationevents at the LTE-edge to resolve any conflicts during thememory replication process for the radio resources.

In 2015, Nunna et al. [79] proposed a real-time context-aware collaboration system by combining MEC with 5Gnetworks. By integrating MEC and 5G, it empowers real-time collaboration systems utilizing context-aware applica-tion platforms. These systems require context informationcombined with geographical information and low latencycommunication. The 4G network might not be capable tofulfill such requirements, instead 5G networks and MEC areproficient to utilize contextual information to provide real-timecollaboration. The above suggested model is beneficial forscenarios such as life Remote Robotic Tele-surgery and RoadAccident that demand high bandwidth and ultra low latency.

In 2016, Kumar et al. [80] proposed a vehicular delay-tolerant network-based smart grid data management scheme.The authors investigated the use of Vehicular Delay-TolerantNetworks (VDTNs) to transmit data to multiple smart griddevices exploring the MEC environment. With the use of astore-and-carry forward mechanism for message transmission,the possible network bottleneck and data latency is avoided.Due to the high mobility of vehicles, a smart grid environmentsupported by MEC is used to monitor large data sets transmit-ted by several smart devices. According to the data movement,these devices make computation charging and dischargingdecisions with respect to message transmission delay, responsetime and high network throughput for movable vehicles.

C. Storage

Limited storage resources of end devices tend to affect userexperience. End-users may utilize MEC resources to overcometheir device storage limitation.

In 2016, Jararweh et al. [12] proposed a Software Definedsystem for Mobile Edge Computing (SDMEC). The proposedframework connects software defined system components toMEC to further extend MCC capabilities. The componentsjointly work cohesively to enhance MCC into the MECservices. Working with Software Defined Networking (SDN),Software Defined Compute (SDCompute), Software DefinedStorage (SDStorage), and Software Defined Security (SDSec)are the prime focus of the proposed framework which enablesapplications requiring compute and storage resources. Appli-cations like traffic monitoring, content sharing and mobilegaming will benefit from SDMEC.

D. Energy Efficiency

As previously mentioned, the MEC architecture is designedto reduce energy consumption of user devices by migratingcompute intensive tasks to the edge of the network.

In 2014, Wei Gao [81] proposed a opportunistic peer-to-peer mobile cloud computing framework. The probabilisticframework is comprised of peer mobile devices that areconnected via their short-range radios. These mobile devicesare enabled to share both the energy and computationalresources depending on their available capacity. He proposedthe probabilistic method to estimate the opportunistic networktransmission status and ensure the resultant computation istimely delivered to its initiator. The purpose of the proposedframework is to facilitate warfighters at the tactical edgein a war zone. This framework is beneficial for situationawareness or surrounded ground environment understanding,with the help of data processed by in-situ (on site) sensors. Thepreambled novel framework thus efficiently shares computa-tional tasks by migrating workloads among warfighters mobilehand held devices, perhaps taking an account of timeliness ofcomputational workload for successive resultant migration.

In 2015, Beck et al. [82] proposed ME-VoLTE which isan architecture that integrates MEC to voice over LTE. Theencoding of video calls is offloaded to the MEC server locatedat the base station (eNodeB). The offloading of video encodingthrough external services helps escalating battery lifetime ofthe user equipment. Encoding is high computational-intensiveand hence is very power consuming. In the proposed system,encoding techniques are wisely used to stream video onthe MEC server. MEC transcodes video by using a specialcodec program before responding to the user device request.This phenomenon significantly increases data transmission andenhances power management.

In 2015, El-Barbary et al. [25] proposed DroidCloudletwhich is based on commodity mobile devices. DroidCloudletis legitimized with resource-rich mobile devices that take theload of resource-constraint mobile devices. The purpose of theproposed architecture is to enhance mobile battery lifetime bymigrating data-intensive and compute-intensive tasks to rich-media. DroidCloulet works as a client device or as a server


device running an application that supplements resource-poordevices by offering their available resources. One of thedevices takes the role of an agent which is responsible forcoordinating resources with other groups of devices.

In 2016 Jalali et al. [83] proposed a flow-based and time-based energy consumption model. They conducted number ofexperiments for efficient energy consumption using centralizednano Data Centers (nDCs) in a cloud computing environment.The authors claim that nDCs energy consumption is not yetbeen investigated. Therefore, several models were presented toperform energy consumption tests on both shared and unsharednetwork equipment. In the paper, it concludes that nDCsmay lead to energy savings if the applications, especially IoTapplications generate and process data within user premises.

E. Summary

In this section, we presented the relevant state-of-the-art re-search results in the MEC area. Among the discussed researchfocuses, computational offloading, low latency and energyefficiency have received more attention by the MEC com-munity. Considering the former, two predominant categoriesof approaches have been proposed to enable computationaloffloading in MEC network models: i) algorithmic solutionssuch as those based on the game theory and successive convexapproximation [74], [75], and ii) network architecture basedsolutions, such as cloud of co-located mobile devices [77].With respect to low latency, early approaches basically rely onthe wireless network technologies for latency reduction, e.g.,through integrating MEC systems with LTE or 5G. Recently,domain-specific network technologies such as vehicular delay-tolerant networks have been used to mitigate the latencyin MEC systems. The state-of-the-art in MEC-based energyefficiency is, in most cases, addressed indirectly through com-putational offloading mechanisms, such as in [82] and [25].Other existing approaches have attempted to address this issuethrough proprietary network architectures, such as peer-to-peermobile networks [81] and centralized nano data centers [83].

VI. RESEARCH ON INFRASTRUCTURE

There are a few contributions on the MEC infrastructurewhich has been partly discussed in the literature [15], [17].In this section, we explore existing MEC infrastructures withrespect to their deployment scenarios and developed test beds.

A. Deployment Scenarios

As mentioned earlier, MEC can be flexibly and intelligentlydeployed at different sites, including UMTS Radio AccessNetwork (UTRAN), LTE E-UTRAN Node B, 3G Radio Net-work Controller (RNC) and multi-Radio Access Technology(RAT). A MEC deployment may use the shared or dedicatedNetwork Functions Virtualization (NFV) architecture.

According to the first release of Information Services Group(ISG) MEC, the implementation scenarios can either be at theoutdoor environment, such as LTE and 3G sites or the indoorenvironment, such as shopping malls and hospitals.

1) MEC in outdoor scenarios: Several ways are possibleto implement MEC in outdoor scenarios, for example,

macro cells vendors insert virtualization environmentinto a radio access network. This scenario helps opera-tors to deliver network features with high value services.Moreover, it improves QoE by providing low latency,pushes more intelligence to the edge and provides bettercomputation offloading. The infrastructure where MECis closely integrated with RAN, gives a better networktraffic analysis, radio network status, and device locationservices.

2) MEC in indoor scenarios: In WiFi or 3G/4G accesspoints, MEC can be deployed through light weightvirtualization. Its deployment in machine-to-machineenvironments can monitor temperature, humidity, airconditioning, etc. with the help of connected sensors atvarious indoor locations. MEC can also be beneficialin case of any emergency situation, such as in anyhazardous situation in a residential building environmentwhere it can help people to evacuate the building withthe help of AR services.

B. MEC Testbeds

This subsection lists some recent testbeds that are developedand tested by implementing mobile edge computing platforms.

1) 5th generation test network: The 5th Generation TestNetwork (5GTN) architecture was developed and successfullytested at Oulu, Finland, which is based on LTE and LTE-Advanced (LTE-A) technology [84]. It opens an opportunityfor application developers to develop their applications in atest environment before they are brought to the market. Theintroduced testbed is composed of different environments, oneis located at Technical Research Centre of Finland (VTT’s) 5Glaboratory and the other one is at the University of Oulu’s Cen-tre for Wireless Communications (CWC). The CWC networkis opened for public users, whereas VTT’s network is in a moresecured and private environment. Both networks are integratedwith the help of carrier-grade technology offering a real-timeenvironment. The private network is connected to 5G testlaboratories located in different parts of Europe. The purposeis to stretch 5G network functionality. The CWC network wastargeted for any mobile user of any mobile operator. The keypurpose is to give access to the university students and visitorswith high-nature 5G experience. MEC functionality is basedon a Nokia provided solution that is operative in an AirFramecloud environment and can be tested in the 5GTN architecture.It will allow the third-parties service providers to test theirapplications in a MEC-5G.

2) Industrial Testbeds: Nokia and China mobile success-fully tested advance mobile solutions for utmost mobile datacapacity and real-time video [85]. The testbed was deployedin a car race stadium where 11707 active users were simulta-neously connected with small cells and 6195 users with macrocells. In total, 95 LTE small cells were installed having 2.6TDD, 2.3 TDD and 1.8 FDD specifications at the ultra-densedistance of 10-15m. The platform was built for MEC withairframe Radio Cloud platform for MEC and Airscale WiFiwith flexi zone controllers. The system successfully deliveredhigh performance HD videos on user mobile panels offering


multi-screen view. Similarly, another testbed application wascreated by Nokia and Chunghwa Telecom (CHT) implementedat a baseball stadium which gives a live TV coverage like viewand live experience of match atmosphere simultaneous at thesame time [86]. The MEC environment was created with thehelp of Nokia Flexi Zone base stations that use 30 MHz ofLTE spectrum. Spectators are able to see four video feeds atthe same time that are on a split mobile screen. MEC offersultra-low latency which is required for live video streamingby moving compute power to process the videos at the nearestplace to the subscribers.

Nokia and its partners delivered an intelligent car-to-carinfrastructure communication system using operators live LTEnetwork [87]. Vehicles connection is facilitated by differentcloudlets deployed on the Nokia MEC platform at mobilebase stations as shown in Figure 5. These cloudlets wereable to deliver end-to-end latency below 20ms. First testeduse case was emergency brake or slowing down car prior toany upcoming emergency. Vehicles can communicate almostin real-time with the vehicles that are even beyond sight.The second tested use case was cooperative passing assistantthat also utilizes cloudlets deployed at the LTE base stations.Vehicles changing lanes are alarmed with the critical distancebetween them. On the basis of distance and car velocity, thesituation is computed by the cloudlets and later signaled tovehicles with guidance of possible actions to avoid any risks.

Fig. 5. Vehicular Communication System

VII. SECURITY AND PRIVACY ISSUES

This section studies security and privacy issues in thecontext of different architectural elements of MEC [88], [89],[90], [91].

A. Security Concerns

The components of CIA triad, Confidentiality, Integrity,and Availability makeup a model design for informationsecurity [92]. There are several aspects of trust that needconsiderations in the MEC infrastructure.

1) Confidentiality: There are several applications hostedat the edge of the network providing their services to

mobile users, e.g., location-awareness. In spite of thefact that these applications are beneficial but they alsopose confidential risks. For example, at the applicationlayer there is no rule defined to separate user identityfrom its geo-location [93]. Therefore, new protocolsare required to be preambled. The user informationis vulnerable between MEC and cloud communicationchannels. Intercepting the communication stream, suchas packet sniffing, will exploit location-based attacks onend devices.

2) Integrity: The MEC ecosystem incorporates multipleactors, such as end-users, service providers, and infras-tructure providers [13]. This causes several security chal-lenges. Cloud servers efficiently enable compute nodesto authenticate them to administrative servers in datacenters due to isolated environments but is less suitablein an open environment. For example, MEC nodes undera multi-management domain will be difficult to sharetheir identification with cloud servers. This scenario cancause several attacks, such as man-in-the-middle attackin which the attacker can authenticate themselves to thecentral cloud systems and later with end devices to stealtheir secret information.

3) Availability: Due to less isolated environment, MECsystem may suffer Denial of Service (DOS) attacks thatcould be application- or packet-based [17]. On a singlenode, these attacks might not be much hazardous but ifthe correlative attacks occur simultaneously at multiplegeo-locations, it can lead to serious implications. For ex-ample, compromised sensors in the industrial sector willmake a ripple effect globally. Such attacks are difficult tomitigate, as MEC systems are directly connected to theend devices and there is no way of detecting maliciousnetwork activity.

1) Network Security: As the preponderance of variouscommunication networks, such as mobile core networks orwireless networks, network security is a very important ele-ment in MEC environments. In the traditional network secu-rity environment, the network administrator defines networksecurity policies which isolate network traffic. However, thedeployment of MEC at the Internet edge, stresses the net-work management policy which may be vulnerable to variousattacks, such as DOS which may damage MEC and causeuseless heavy network traffic. This kind of attack is limited toMEC nodes and not much effective to the back-haul networksince the back-haul network is more secured. Attackers canalso launch traffic injection or eavesdropping attacks that cantakeover the network or a quantum of a network. Hackershijacking the network stream can launch attacks to affect MECsystem performance. Man-in-the-middle attack is likely to beeffective when intercepting data communication. Attackers cansuccessfully manipulate the data traveling from the cloud tothe user and vice versa. It is difficult to mitigate such attacksbecause deploying and dropping virtual machines make itcumbersome to maintain the blacklist. Han et al. [94] proposeda measurement-based approach to prevent user connectionwith rogue gateways by observing the round-trip time between


the user and the DNS server.2) Core Network Security: It should be noted that all edge

paradigms may be supported by the core network and most ofthe core network security is enabled by mobile core networksor a central cloud. The security of cloud services is mostlymanaged by third party suppliers, such as Amazon, Microsoftand Google. However, it is not possible to completely rely ontheir security mechanisms. In addition, there is a high risk foruser’s personal and sensitive information which can be stolenby malicious entities. Edge devices exchange information witheach other and may bypass the central system’s securitymechanism. This makes security vulnerable and hackable. Thistype of security issue will not affect the whole ecosystem andwill be limited due to its decentralized nature. There is alsoa possibility of the system data to be changed and providefalse information, if the services are hijacked. The level ofthis effect will be limited but may cause denial of services.If the core infrastructure is compromised then it can sabotagesome elements of the core system. Core network elements thatare compromised can disrupt the lower level infrastructure.Attackers may have full access to the information and maytamper the network data flow.

3) MEC Server Security: MEC at the edge comprises ofseveral virtualized servers providing intelligent IT services.However, these services are liable to external security threats,for example, physical access to the data centre is less protectedor guarded. Attackers breaching security channels can physi-cally damage IT resources. This particular attack is limited toa specific geographical location and may not be very critical.Moreover, stream of information to and from the local scopeof data centre can be stolen from malicious actors, such asusers and ASPs. In addition, design flaws, configuration errors,insufficient security training or abusing one’s own privilegesmay be an alarming risk to the security of data center systems.Being newly preambled in the technology world, MEC lackssome security expertise for adequate system security. Once thelogin access has been granted to the MEC system resources,the attacker can abuse system integrity or can execute denialof service attacks, man-in-the-middle attacks, etc. The servicesare discontinued or interrupted as a result of such securitybreaches. Another security issue is the compromise of anentire data centre. In this type of attack, the whole data centeris hijacked through a single or a combination of differentattacks. The attacks might be privilege escalation or a fakeinfrastructure. A compromised data center has a large impactover geographical locations which may result in high scaledamage.

4) Virtualization Security: In core mobile edge data centers,several network instances co-exist sharing network instances.If one resource is compromised, it can affect the whole virtual-ized infrastructure. An attacker may misuse and exploit systemresources that have been conceded. Denial of Service (DOS)attacks are most likely to happen. MEC virtualized systemscan completely drain the resources that serve computational,storage and network tasks. Users connecting to MEC virtualservers may face denial of requests and services. Furthermore,malicious antagonists can misuse virtual resources and notonly affect the system itself but also IoT devices that are

connected to it. For instance, any IoT device that is in therange of the radio network and is vulnerable can be hacked andsabotaged. One of the common security concerns is privacyleakage. Several APIs implemented in the MEC environmentare responsible to deliver information of the physical andlogical infrastructure. However, these APIs are most likelyto be less protected against any malevolent activity. Severalattacks can be escalated, e.g., malicious virtual machineshosted in a data center can advance to other virtual machines orother data centers. Users moving across different geographicallocations can escalate such attacks to other MEC virtualizedservers. The virtual machine itself, affected by an attacker, canbecome a hostile and launch attacks on other virtual machineshosted on the same system.

5) End Devices Security: End-user devices can potentiallybe harmful to affect some elements of the ecosystem. How-ever, the impact could be narrow due to limited user devicesurroundings. User devices act as a carrier in a distributedenvironment. In addition, there could be rogue users that canintrude the system and perform some malicious activities. Forexample, they can inject false values or information to thesystem. A device can be reconfigured and set to send fakeinformation, such as wrong surveillance camera information orincorrect data announcements by vehicles. Moreover, there aresome scenarios where devices can participate in service ma-nipulation. For example, any compromised device connectedin a cluster environment can change and control services inthat cluster.

B. Security MechanismsSecurity breaches may cause potential harmful problems

within the system. Therefore, it is very important to implementappropriate security mechanisms and safeguard the MECresources from any intrusion. In this subsection, we presentexisting security mechanisms for MEC.

1) Identification and Authentication: In a cloud computingenvironment, data centers are mostly hosted by cloud serviceproviders, whereas in all edge paradigms, providers may behosted by several providers depending on their choices. Forexample, cloud service providers may extend their IT servicesto the edge using existing infrastructures. MEC resourceproviders may differ with the extended cloud infrastructureand the end-user may want to use limited cloud resourcesdepending on their budget and lease their resources in thelocal cloud. In order to integrate all these services, properidentification and authentication is required. Every entity inthe ecosystem, such as end devices, virtual machine services,the cloud and MEC infrastructure service providers, andapplication service providers should be able to identify andmutually authenticate each other.

In [95], a user-friendly solution is introduced to providea secure authentication in a local ad-hoc wireless network.The connected devices share only limited public informationthat enables them to exchange authenticated key protocols.Similarly, NFC also enables a secure authentication method ina cloudlet scenario [96]. NFC applications based on cloudletsenable authentication by NFC-equipped end devices. More-over, it is also of prime importance to have a continuous


connectivity of user devices with their respected cloud servers.Stand-Alone authentication is introduced for a scenario ifthere is temporary disconnection between MEC and the cloudserver [90]. If the connection is fragile then Stand-Aloneauthentication would be able to authenticate users with thecloud servers. With the evolution of biometric authentication,such as face recognition, eye recognition, and finger print,it will be very helpful to introduce biometric authenticationsystems in MEC.

2) Network Security Mechanisms: Network security is oneof the prime concern for the MEC concept due to the predom-inance of the network infrastructure. Attackers are involvedmore in launching attacks, such as man-in-the-middle andDOS to sabotage the network environment. Thus, it is very cru-cial to deploy a comprehensive security mechanism. Intrusiondetection and prevention is an important prerequisite beforedesigning the MEC infrastructure. Several network entitiescould be vulnerable to any threats, hence these entities shouldbe monitored from internal or external threats. In such cases,the edge infrastructure should be in charge of monitoringits network and equally coordinate with surrounded and corenetworks. Intrusion Detection Systems (IDS) can be employedin the MEC data center to monitor and analyze system logsfor any unauthorized access. It can also be employed at theMEC network side to detect and prevent the network fromany attacks, such as man-in-the-middle attack, DOS and portscanning. A Cloudlet that is located at one hop away frommobile devices can efficiently be meshed to form a securityframework to detect any intrusion [97]. Cloudlets can serve asa proxy for distant cloud servers in case of any unavailabilityissue due to certain attacks. Moreover, by implementing SDN,it is easy to reduce network cost and scale network resources.SDN can isolate network traffic and segregate malicious data.A proposed access control scheme based on OpenFlow isuseful for multiple security requirements [98]. For example,having direct access to the network components will makeit easier to monitor and detect any abnormal activity in thenetwork traffic.

3) Virtualization Security Mechanisms: Virtualization tech-nology is one of the foundation for the edge paradigm andthus it is of paramount importance to be secured. Maliciouselements that get access to virtual servers may hijack the entireedge data center. Virtualized servers and their hosted physicalservers can be protected through hypervisor hardening, net-work abstractions, and isolation policies [99].

4) Data Security: In an edge paradigm, user data isoutsourced to the MEC server that gives access control tomobile users. This introduces some challenges, such as dataintegrity and authorization, for example, outsourced data canbe modified or disappeared, and uploaded data can be accessedfor malicious activity. Moreover, data owners and data serverspossess dissimilar identities and business interests that makethe scenario more vulnerable. Appropriate auditing methodscan be used to audit data storage to confirm that data isproperly resided in the cloud [100].

5) Data Computation Security: Securing data computationis another important issue that has to be addressed. Thereare two major aspects to secure any computation that is

outsourced, including computation verification and data en-cryption. Verifiable computing allows the computing node tooffload some functions to other servers that could not betrusted, but it enables the maintenance of the results thatare verifiable. Other servers perform a check on the givenfunction and confirm the correctness of computation. Thereshould be a mechanism through which the user is allowed toverify computational accuracy. In [101], a verifiable computingprotocol is proposed to return a computational-sound, non-interactive proof. Data encryption is another security mech-anism. The data that is sent from user devices need to beprotected and encrypted before outsourced to MEC servers.One of the popular services is keyword search which meansto search keywords from encrypted data files. In [102], astatistical measuring approach is proposed to search througha secured searchable index. The index is secured through aone-to-many order-preserving mapping approach.

C. Privacy IssuesDue to close proximity to the end-user, privacy preserving,

such as data, usage and location may be challenging in mobileedge computing. User privacy breaches may become worse ifthe attacker gains personal information, such as the credit cardinformation, personal emails, etc. Data privacy or informationprivacy of the user have a risk of being accessed. It could beeven worse if the attacker gets access to the user’s sensitiveinformation.

Aggregation schemes, such as homomorphic encryption canenable privacy-preserving aggregation at gateways to secureuser information [103]. An attacker may retrieve user in-formation by learning the usage pattern of the user devicewhile accessing MEC. For example, in a home smart-gridenvironment, meter reading, such as the presence and absenceof the user at home and user in-house behavior can helpthe attacker to perform any malicious or criminal activities.Non-Intrusive Load Leveling (NILL) has been introduced toencounter these types of issues [104]. However, it cannot beimplemented in MEC environments due to untrusted thirdparties, e.g., the lack of a duplicate battery to perform NILL.One possible way to counter this kind of privacy is creatingdummy tasks and performing multiple offloadings to differentlocations. This can therefore hide the original tasks by hidingbehind these fake ones.

Another privacy issue is the user location. The GlobalPositioning System (GPS) is very useful for users to benefitfrom geo-location services. Mobile users use location-basedservices for GPS navigation. However, such services endurecertain privacy issues, for example, users who share theirlocation information with location-based services. The userdevice connected to the nearest MEC will give an indicationto the attacker that which MEC location the compute deviceis near to, as shown in Figure 6. In order to secure locationinformation, there are several ways to confuse the attacker.For example, the MobiShare system is a flexible and securemechanism for sharing geo-location information and it has agood support for location-based applications [105].

Moreover, the edge paradigm in general, and MEC inparticular can be used to improve the privacy level of certain


Fig. 6. Archtecture of Location Sharing

services, such as crowd sourcing. The state-of-the-art pri-vacy solutions are not particularly suitable for crowd sourcedlocation-based services. Abdo et al. [106] proposed a solutionto deploy a crowd-sourcing platform in a trusted edge datacenter to protect the anonymity of the participants of certainlocation-based services.

VIII. OTHER OPEN ISSUES

As a recent technology platform, not much research hasbeen specifically advocated to mobile edge computing. Thereare some issues in MEC that need to be addressed before itscommercial deployments. This section discusses and identifiesthe open issues that are investigated by different researchersin the development of MEC.

A. Security

Security is one of the main concerns for technology ad-visers to secure MEC deployments. There are some securitymechanisms that are applicable to MEC, as discussed in theprevious section. However, there are still some issues thatneed proper research studies. For example, computational-intensive applications outsource their computation to MECservers. Computation tasks are performed through wirelessmedium opening up the risk of intrusion. Moreover, differentusers connected to the common physical server also raisesecurity issues [107]. The application data movement is possi-ble through encryption and decryption strategies but it affectsapplication performance.

B. Resource Optimization

Promoting cloud infrastructures to the network edge, MECincorporates less resources than the cloud. Computationaloffloading is supported by applications and virtualized MECservers. However, computational tasks carry extra overloaddue to heterogeneous processor architectures. For example,mobile smart phones have mostly ARM and x86 architectures,hence they need translation or emulation [108]. Therefore,an optimized solution for enhancing performance of intrinsiclimited resources is required [109].

C. Transparent Application MigrationAs mentioned previously, user applications are transported

to MEC servers for execution. It is very vital to transparentlymigrate these applications for usability of delay-sensitive mo-bile applications, such as real-time applications [110]. Poorcompute performance and delay in service provisioning deteri-orate the emergence of mobile applications [111]. Applicationmigration is a software level solution that can be achieved byproposing solutions to optimize cloud resources utilization atthe edge [112].

D. PricingMobile edge computing environments involve several actors

that quote different prices for their services. These actors havedifferent payment methods, different customer managementmodels, and different business policies. Therefore, it raisesseveral questions: i) what will be the mutually agreed price, ii)what will be the mode of payment, and iii) who will processcustomer payment. For example, a game application on theuser device has to utilize cloud resources, the mobile networkand game services. The user has to pay for the game whichshould be divided equally or as per mutual contract to allthe entities involved. This can be argued that agreeing to thepricing may be difficult among different entities.

E. Web InterfaceCurrently, the interface used to access MEC and the cloud

is the web interface which is not sufficient for user experience.The web interface is generally not designed for mobile devicesand hence have compatibility issues. Generally, web interfacesare not suitable for MEC due to their overhead problem. There-fore, standard protocols are required for smooth communica-tion between the user, MEC and the cloud. The latest versionof HTML5 is designed specifically for advanced devices, suchas mobile or smart phones. However, performance and test-based research is needed to make HTML5 ready for MEC.

F. Other IssuesMany issues are already discussed in previous sections, but

in addition, there are some other issues which are also vitalto strengthen the MEC framework. A comprehensive scientificresearch study is required to avoid any security issues that candamage the system.

• Openness of the Network: Mobile networks have com-plete authority over the network but in the case of MECit will be very crucial to open the network for third partyvendors due to possible security risks.

• Multiservices and Operations: ASPs, OTT, network ven-dors and content providers require access to MEC datacenters. This scenario causes complexity for seamlessintegration with third-party services.

• Robustness and Resilience: Deploying resources at MECis very important to enable robustness of MEC servers.

• Security and Privacy: User privacy and its data securitymay be exposed while integrating mobile services withMEC. Prior to the MEC deployment, there should be anassurance that the infrastructure is well protected.


IX. CONCLUSIONS

Mobile edge computing has a great potential to be the futureedge technology offering bandwidth, battery life and storage tothe resource-constraint mobile devices. MEC trends to provideelastic resources at the end of the network towards appli-cations demanding computational-intensive tasks with highbandwidth and ultra low latency, especially in 5G networks.MEC deployment can build an ecosystem involving third-party partners, content providers, application developers, OTTplayers, network vendors and multiple mobile operators. Thispaper has presented a thorough study on the recent researchand technological development in the area of MEC and itsapplication domains, research challenges and open issues.

REFERENCES

[1] E. Cau, M. Corici, P. Bellavista, L. Foschini, G. Carella, A. Edmonds,and T. M. Bohnert, “Efficient Exploitation of Mobile Edge Com-puting for Virtualized 5G in EPC Architectures,” in 2016 4th IEEEInternational Conference on Mobile Cloud Computing, Services, andEngineering (MobileCloud), March 2016, pp. 100–109.

[2] C. Systems, “Cisco visual networking index: Globalmobile data traffic forecast update, 20152020 white paper,”http://www.cisco.com/c/en/us/solutions/collateral/service-provider/visual-networking-index-vni/mobile-white-paper-c11-520862.html,accessed August 22, 2016.

[3] G. Orsini, D. Bade, and W. Lamersdorf, “Computing at the mobileedge: Designing elastic android applications for computation offload-ing,” in 2015 8th IFIP Wireless and Mobile Networking Conference(WMNC), Oct 2015, pp. 112–119.

[4] E. Borgia, R. Bruno, M. Conti, D. Mascitti, and A. Passarella, “Mo-bile edge clouds for information-centric iot services,” in 2016 IEEESymposium on Computers and Communication (ISCC), June 2016, pp.422–428.

[5] M. A. Marotta, L. R. Faganello, M. A. K. Schimuneck, L. Z. Granville,J. Rochol, and C. B. Both, “Managing mobile cloud computingconsidering objective and subjective perspectives,” Computer Networks,vol. 93, Part 3, pp. 531 – 542, 2015, cloud Networking andCommunications {II}. [Online]. Available: http://www.sciencedirect.com/science/article/pii/S1389128615003667

[6] H. T. Dinh, C. Lee, D. Niyato, and P. Wang, “A survey of mobilecloud computing: architecture, applications, and approaches,” WirelessCommunications and Mobile Computing, vol. 13, no. 18, pp. 1587–1611, 2013. [Online]. Available: http://dx.doi.org/10.1002/wcm.1203

[7] Y. Jararweh, A. Doulat, O. AlQudah, E. Ahmed, M. Al-Ayyoub, andE. Benkhelifa, “The future of mobile cloud computing: Integratingcloudlets and mobile edge computing,” in 2016 23rd InternationalConference on Telecommunications (ICT), May 2016, pp. 1–5.

[8] S. Kitanov, E. Monteiro, and T. Janevski, “5g and the fog 2014;survey of related technologies and research directions,” in 2016 18thMediterranean Electrotechnical Conference (MELECON), April 2016,pp. 1–6.

[9] M. T. Beck, M. Werner, S. Feld, and S. Schimper, “Mobile edgecomputing: A taxonomy,” in Proc. of the Sixth International Conferenceon Advances in Future Internet. Citeseer, 2014.

[10] K. Kai, W. Cong, and L. Tao, “Fog computing for vehicular ad-hocnetworks: paradigms, scenarios, and issues,” The Journal of ChinaUniversities of Posts and Telecommunications, vol. 23, no. 2, pp. 56–96, 2016.

[11] S. Yi, C. Li, and Q. Li, “A survey of fog computing: Concepts,applications and issues,” in Proceedings of the 2015 Workshopon Mobile Big Data, ser. Mobidata ’15. New York, NY, USA:ACM, 2015, pp. 37–42. [Online]. Available: http://doi.acm.org/10.1145/2757384.2757397

[12] Y. Jararweh, A. Doulat, A. Darabseh, M. Alsmirat, M. Al-Ayyoub,and E. Benkhelifa, “Sdmec: Software defined system for mobileedge computing,” in 2016 IEEE International Conference on CloudEngineering Workshop (IC2EW), April 2016, pp. 88–93.

[13] R. Roman, J. Lopez, and M. Mambo, “Mobile edge computing, fog etal.: A survey and analysis of security threats and challenges,” FutureGeneration Computer Systems, pp. –, 2016. [Online]. Available:http://www.sciencedirect.com/science/article/pii/S0167739X16305635

[14] A. Ahmed and E. Ahmed, “A survey on mobile edge computing,” in2016 10th International Conference on Intelligent Systems and Control(ISCO), 2016, pp. 1–8.

[15] Y. C. Hu, M. Patel, D. Sabella, N. Sprecher, and V. Young, “Mobileedge computinga key technology towards 5g,” ETSI White Paper,vol. 11, 2015.

[16] D. Satria, D. Park, and M. Jo, “Recovery for overloaded mobileedge computing,” Future Generation Computer Systems, pp. –, 2016.[Online]. Available: http://www.sciencedirect.com/science/article/pii/S0167739X16302096

[17] M. Patel, B. Naughton, C. Chan, N. Sprecher, S. Abeta, A. Neal et al.,“Mobile-edge computing introductory technical white paper,” WhitePaper, Mobile-edge Computing (MEC) industry initiative, 2014.

[18] S. Yi, Z. Hao, Z. Qin, and Q. Li, “Fog computing: Platform and ap-plications,” in Hot Topics in Web Systems and Technologies (HotWeb),2015 Third IEEE Workshop on. IEEE, 2015, pp. 73–78.

[19] P. Asrani, “Mobile cloud computing,” International Journal of Engi-neering and Advanced Technology (IJEAT), vol. 2, no. 4, pp. 606–609,2013.

[20] D. Huang et al., “Mobile cloud computing,” IEEE COMSOC Multime-dia Communications Technical Committee (MMTC) E-Letter, vol. 6,no. 10, pp. 27–31, 2011.

[21] P. Mell and T. Grance, “The nist definition of cloud computing,” 2011.[22] T. Brummett, P. Sheinidashtegol, D. Sarkar, and M. Galloway, “Per-

formance metrics of local cloud computing architectures,” in CyberSecurity and Cloud Computing (CSCloud), 2015 IEEE 2nd Interna-tional Conference on, Nov 2015, pp. 25–30.

[23] T. Zhao, S. Zhou, X. Guo, Y. Zhao, and Z. Niu, “A cooperativescheduling scheme of local cloud and internet cloud for delay-awaremobile cloud computing,” in 2015 IEEE Globecom Workshops (GCWkshps), Dec 2015, pp. 1–6.

[24] Y. Liu, M. J. Lee, and Y. Zheng, “Adaptive multi-resource allocationfor cloudlet-based mobile cloud computing system,” IEEE Transactionson Mobile Computing, vol. 15, no. 10, pp. 2398–2410, Oct 2016.

[25] A. E. H. G. El-Barbary, L. A. A. El-Sayed, H. H. Aly, and M. N.El-Derini, “A cloudlet architecture using mobile devices,” in 2015IEEE/ACS 12th International Conference of Computer Systems andApplications (AICCSA), Nov 2015, pp. 1–8.

[26] T. Verbelen, P. Simoens, F. De Turck, and B. Dhoedt, “Cloudlets:Bringing the cloud to the mobile user,” in Proceedings of the ThirdACM Workshop on Mobile Cloud Computing and Services, ser.MCS ’12. New York, NY, USA: ACM, 2012, pp. 29–36. [Online].Available: http://doi.acm.org/10.1145/2307849.2307858

[27] Z. Pang, L. Sun, Z. Wang, E. Tian, and S. Yang, “A survey of cloudletbased mobile computing,” in 2015 International Conference on CloudComputing and Big Data (CCBD), Nov 2015, pp. 268–275.

[28] A. Taherkordi and F. Eliassen, “Poster abstract: Data-centric iot servicesprovisioning in fog-cloud computing systems,” in 2017 IEEE/ACMSecond International Conference on Internet-of-Things Design andImplementation (IoTDI), 2017.

[29] S. S. Panwar, “Heterogeneous cellular networks.”[30] A. H. Khan, M. A. Qadeer, J. A. Ansari, and S. Waheed, “4g as a next

generation wireless network,” in Future Computer and Communication,2009. ICFCC 2009. International Conference on, April 2009, pp. 334–338.

[31] CommVerge, “Radio access network (ran) optimization,” 2016,[Online; accessed 19-September-2016]. [Online]. Available: \url{http://www.commverge.com/Solutions/SubscribersServicesManagement/RANOptimization/tabid/174/Default.aspx}

[32] C.-K. Park, “Performance for radio access network in mobile backhaulnetwork,” The Journal of The Institute of Internet, Broadcasting andCommunication, vol. 12, no. 6, pp. 297–302, 2012.

[33] Brocade, “Brocade and the mobile edge,” 2016,[Online; accessed 03-October-2016]. [Online]. Avail-able: \url{https://www.brocade.com/content/dam/common/documents/content-types/solution-brief/brocade-and-the-mobile-edge-sb.pdf}

[34] J. Wu, Z. Zhang, Y. Hong, and Y. Wen, “Cloud radio access network(c-ran): a primer,” IEEE Network, vol. 29, no. 1, pp. 35–41, 2015.

[35] T. H. Luan, L. Gao, Z. Li, Y. Xiang, and L. Sun, “Fog computing: Fo-cusing on mobile users at the edge,” arXiv preprint arXiv:1502.01815,2015.

[36] N. Zhong, “Mobile edge computing: Unleashing thevalue chain,” 2015, [Online; accessed 07-October-2016].[Online]. Available: \url{http://dashif.org/wp-content/uploads/2015/08/6d-Mobile-Edge-Computing-Unleashing-the-value-chain.pdf}


[37] J. Sharpe, “How mobile edge computing is helpingoperators face the challenges of todays evolving mobilenetworks,” 2015, [Online; accessed 03-October-2016].[Online]. Available: \url{http://eecatalog.com/intel/2015/08/17/how-mobile-edge-computing-is-helping-operators-face-the-challe\-nges-of-todays-evolving-mobile-networks/}

[38] A. V. Dastjerdi, H. Gupta, R. N. Calheiros, S. K. Ghosh, and R. Buyya,“Fog computing: Principals, architectures, and applications,” arXivpreprint arXiv:1601.02752, 2016.

[39] T. Olsson and M. Salo, “Online user survey on current mobile aug-mented reality applications,” in Mixed and Augmented Reality (ISMAR),2011 10th IEEE International Symposium on, Oct 2011, pp. 75–84.

[40] T. Piumsomboon, A. Clark, M. Billinghurst, and A. Cockburn,“User-defined gestures for augmented reality,” in CHI ’13 ExtendedAbstracts on Human Factors in Computing Systems, ser. CHI EA ’13.New York, NY, USA: ACM, 2013, pp. 955–960. [Online]. Available:http://doi.acm.org/10.1145/2468356.2468527

[41] R. Azuma, Y. Baillot, R. Behringer, S. Feiner, S. Julier, and B. MacIn-tyre, “Recent advances in augmented reality,” IEEE Computer Graphicsand Applications, vol. 21, no. 6, pp. 34–47, Nov 2001.

[42] S. Yuen, G. Yaoyuneyong, and E. Johnson, “Augmented reality: Anoverview and five directions for ar in education,” Journal of Educa-tional Technology Development and Exchange, vol. 4, no. 1, pp. 119–140, 2011.

[43] R. Buyya and A. V. Dastjerdi, Internet of Things: Principles andParadigms. Elsevier, 2016.

[44] Y. Wang, T. Uehara, and R. Sasaki, “Fog computing: Issues andchallenges in security and forensics,” in Computer Software andApplications Conference (COMPSAC), 2015 IEEE 39th Annual, vol. 3.IEEE, 2015, pp. 53–59.

[45] N. Peter, “Fog computing and its real time applications,” 2015.[46] I. Stojmenovic and S. Wen, “The fog computing paradigm: Scenarios

and security issues,” in Computer Science and Information Systems(FedCSIS), 2014 Federated Conference on. IEEE, 2014, pp. 1–8.

[47] E. Ahmed and M. H. Rehmani, “Mobile edge computing: Opportuni-ties, solutions, and challenges,” Future Generation Computer Systems,2016.

[48] Y. Cao, S. Chen, P. Hou, and D. Brown, “Fast: A fog computing assisteddistributed analytics system to monitor fall for stroke mitigation,” inNetworking, Architecture and Storage (NAS), 2015 IEEE InternationalConference on, Aug 2015, pp. 2–11.

[49] A. S. Go, D. Mozaffarian, V. L. Roger, E. J. Benjamin, J. D. Berry,M. J. Blaha, S. Dai, E. S. Ford, C. S. Fox, S. Franco et al., “Heartdisease and stroke statistics-2014 update,” Circulation, vol. 129, no. 3,2014.

[50] P. A. Heidenreich, J. G. Trogdon, O. A. Khavjou, J. Butler, K. Dracup,M. D. Ezekowitz, E. A. Finkelstein, Y. Hong, S. C. Johnston, A. Kheraet al., “Forecasting the future of cardiovascular disease in the unitedstates a policy statement from the american heart association,” Circu-lation, vol. 123, no. 8, pp. 933–944, 2011.

[51] V. Stantchev, A. Barnawi, S. Ghulam, J. Schubert, and G. Tamm,“Smart items, fog and cloud computing as enablers of servitizationin healthcare,” Sensors & Transducers, vol. 185, no. 2, p. 121, 2015.

[52] K. Hong, D. Lillethun, U. Ramachandran, B. Ottenwalder, andB. Koldehofe, “Mobile fog: A programming model for large-scaleapplications on the internet of things,” in Proceedings of the SecondACM SIGCOMM Workshop on Mobile Cloud Computing, ser. MCC’13. New York, NY, USA: ACM, 2013, pp. 15–20. [Online].Available: http://doi.acm.org/10.1145/2491266.2491270

[53] I. International, “Computer technology increasingly aidstraffic management,” 2009, [Online; accessed 10-September-2016]. [Online]. Available: http://www.itsinternational.com/categories/detection-monitoring-machine-vision/features/computer-technology-increasingly-aids-traffic-management/

[54] J. K. Laurila, D. Gatica-Perez, I. Aad, O. Bornet, T.-M.-T. Do,O. Dousse, J. Eberle, M. Miettinen et al., “The mobile data challenge:Big data for mobile computing research,” in Pervasive Computing, no.EPFL-CONF-192489, 2012.

[55] Wikipedia, “Big data — wikipedia, the free encyclopedia,” 2016,[Online; accessed 12-September-2016]. [Online]. Available: \url{https://en.wikipedia.org/w/index.php?title=Big data&oldid=739099684}

[56] R. Misra, B. Panda, and M. Tiwary, “Big data and ict applications:A study,” in Proceedings of the Second International Conferenceon Information and Communication Technology for CompetitiveStrategies, ser. ICTCS ’16. New York, NY, USA: ACM, 2016, pp.41:1–41:6. [Online]. Available: http://doi.acm.org/10.1145/2905055.2905099

[57] TDWI, “Big data analytics,” 2011, [Online; accessed 12-September-2016]. [Online]. Available: \url{https://tdwi.org/portals/big-data-analytics.aspx}

[58] A. Al-Fuqaha, M. Guizani, M. Mohammadi, M. Aledhari, andM. Ayyash, “Internet of things: A survey on enabling technologies,protocols, and applications,” IEEE Communications Surveys Tutorials,vol. 17, no. 4, pp. 2347–2376, Fourthquarter 2015.

[59] M. A. Alsheikh, D. Niyato, S. Lin, H.-P. Tan, and Z. Han, “Mobile bigdata analytics using deep learning and apache spark,” IEEE Network,vol. 30, no. 3, pp. 22–29, 2016.

[60] P. Russom et al., “Big data analytics,” TDWI Best Practices Report,Fourth Quarter, pp. 1–35, 2011.

[61] A. Taherkordi, F. Eliassen, and G. Horn, “From iot big data to iot bigservices,” in Proceedings of the Symposium on Applied Computing, ser.SAC ’17. ACM, 2017.

[62] S. K. Datta, C. Bonnet, and J. Haerri, “Fog computing architectureto enable consumer centric internet of things services,” in 2015International Symposium on Consumer Electronics (ISCE). IEEE,2015, pp. 1–2.

[63] R. Mahmud, R. Vallakati, A. Mukherjee, P. Ranganathan, and A. Nejad-pak, “A survey on smart grid metering infrastructures: Threats and solu-tions,” in 2015 IEEE International Conference on Electro/InformationTechnology (EIT), May 2015, pp. 386–391.

[64] P. Kuakowski, E. Calle, and J. L. Marzo, “Performance studyof wireless sensor and actuator networks in forest fire scenarios,”International Journal of Communication Systems, vol. 26, no. 4, pp.515–529, 2013. [Online]. Available: http://dx.doi.org/10.1002/dac.2311

[65] S. Shin and G. Gu, “Attacking software-defined networks: A firstfeasibility study,” in Proceedings of the Second ACM SIGCOMMWorkshop on Hot Topics in Software Defined Networking, ser. HotSDN’13. New York, NY, USA: ACM, 2013, pp. 165–166. [Online].Available: http://doi.acm.org/10.1145/2491185.2491220

[66] R. Jain and S. Paul, “Network virtualization and software definednetworking for cloud computing: a survey,” IEEE CommunicationsMagazine, vol. 51, no. 11, pp. 24–31, November 2013.

[67] X. Ma, Y. Zhao, L. Zhang, H. Wang, and L. Peng, “When mobileterminals meet the cloud: computation offloading as the bridge,” IEEENetwork, vol. 27, no. 5, pp. 28–33, September 2013.

[68] K. Kumar, J. Liu, Y.-H. Lu, and B. Bhargava, “A survey ofcomputation offloading for mobile systems,” Mobile Networks andApplications, vol. 18, no. 1, pp. 129–140, 2013. [Online]. Available:http://dx.doi.org/10.1007/s11036-012-0368-0

[69] D. Kovachev, T. Yu, and R. Klamma, “Adaptive computation offloadingfrom mobile devices into the cloud,” in 2012 IEEE 10th InternationalSymposium on Parallel and Distributed Processing with Applications,July 2012, pp. 784–791.

[70] M. A. Hassan, M. Xiao, Q. Wei, and S. Chen, “Help your mobileapplications with fog computing,” in Sensing, Communication, andNetworking - Workshops (SECON Workshops), 2015 12th Annual IEEEInternational Conference on, June 2015, pp. 1–6.

[71] M. Satyanarayanan, P. Bahl, R. Caceres, and N. Davies, “The case forvm-based cloudlets in mobile computing,” IEEE Pervasive Computing,vol. 8, no. 4, pp. 14–23, Oct 2009.

[72] K. Kumar and Y.-H. Lu, “Cloud computing for mobile users: Canoffloading computation save energy?” Computer, vol. 43, no. 4, pp.51–56, 2010.

[73] N. Takahashi, H. Tanaka, and R. Kawamura, “Analysis of process as-signment in multi-tier mobile cloud computing and application to edgeaccelerated web browsing,” in Mobile Cloud Computing, Services, andEngineering (MobileCloud), 2015 3rd IEEE International Conferenceon, March 2015, pp. 233–234.

[74] X. Chen, L. Jiao, W. Li, and X. Fu, “Efficient multi-user computationoffloading for mobile-edge cloud computing,” IEEE/ACM Transactionson Networking, vol. 24, no. 5, pp. 2795–2808, Oct 2016.

[75] S. Sardellitti, G. Scutari, and S. Barbarossa, “Joint optimization ofradio and computational resources for multicell mobile-edge comput-ing,” IEEE Transactions on Signal and Information Processing overNetworks, vol. 1, no. 2, pp. 89–103, June 2015.

[76] K. Zhang, Y. Mao, S. Leng, A. Vinel, and Y. Zhang, “Delay constrainedoffloading for mobile edge computing in cloud-enabled vehicularnetworks,” in 2016 8th International Workshop on Resilient NetworksDesign and Modeling (RNDM), Sept 2016, pp. 288–294.

[77] K. Habak, M. Ammar, K. A. Harras, and E. Zegura, “Femto clouds:Leveraging mobile devices to provide cloud service at the edge,” in2015 IEEE 8th International Conference on Cloud Computing, June2015, pp. 9–16.


[78] S. Abdelwahab, B. Hamdaoui, M. Guizani, and T. Znati, “Replisom:Disciplined tiny memory replication for massive iot devices in lte edgecloud,” IEEE Internet of Things Journal, vol. 3, no. 3, pp. 327–338,June 2016.

[79] S. Nunna, A. Kousaridas, M. Ibrahim, M. Dillinger, C. Thuemmler,H. Feussner, and A. Schneider, “Enabling real-time context-awarecollaboration through 5g and mobile edge computing,” in Informa-tion Technology - New Generations (ITNG), 2015 12th InternationalConference on, April 2015, pp. 601–605.

[80] N. Kumar, S. Zeadally, and J. J. P. C. Rodrigues, “Vehicular delay-tolerant networks for smart grid data management using mobile edgecomputing,” IEEE Communications Magazine, vol. 54, no. 10, pp. 60–66, October 2016.

[81] W. Gao, “Opportunistic peer-to-peer mobile cloud computing at thetactical edge,” in 2014 IEEE Military Communications Conference,Oct 2014, pp. 1614–1620.

[82] M. T. Beck, S. Feld, A. Fichtner, C. Linnhoff-Popien, and T. Schimper,“Me-volte: Network functions for energy-efficient video transcoding atthe mobile edge,” in Intelligence in Next Generation Networks (ICIN),2015 18th International Conference on, Feb 2015, pp. 38–44.

[83] F. Jalali, K. Hinton, R. Ayre, T. Alpcan, and R. S. Tucker, “Fogcomputing may help to save energy in cloud computing,” IEEE Journalon Selected Areas in Communications, vol. 34, no. 5, pp. 1728–1739,May 2016.

[84] E. Piri, P. Ruuska, T. Kanstren, J. Makela, J. Korva, A. Hekkala,A. Pouttu, O. Liinamaa, M. Latva-aho, K. Vierimaa et al., “5gtn: Atest network for 5g application development and testing,” in Networksand Communications (EuCNC), 2016 European Conference on. IEEE,2016, pp. 313–318.

[85] NOKIA, “Small cells deliver cost-effective capacity and coverage,indoors and outdoors,” 2016, [Online; accessed 27-November-2016]. [Online]. Available: \url{https://networks.nokia.com/products/small-cells}

[86] ——, “Small cells mobile edge computing coverall the bases for taiwan baseball fans,” 2016,[Online; accessed 27-November-2016]. [Online]. Available:\url{https://blog.networks.nokia.com/small-cells/2016/09/14/small-cells-mobile-edge-computing-cover-bases-taiwan-baseball-fans/}

[87] ——, “Connected cars use case for mobile edge computing,”2016, [Online; accessed 27-November-2016]. [Online]. Available:\url{https://networks.nokia.com/solutions/mobile-edge-computing}

[88] R. Roman, J. Lopez, and M. Mambo, “Mobile edge computing, foget al.: A survey and analysis of security threats and challenges,” arXivpreprint arXiv:1602.00484, 2016.

[89] S. Yi, Z. Qin, and Q. Li, “Security and privacy issues of fog comput-ing: A survey,” in International Conference on Wireless Algorithms,Systems, and Applications. Springer, 2015, pp. 685–695.

[90] I. Stojmenovic, S. Wen, X. Huang, and H. Luan, “An overview offog computing and its security issues,” Concurrency and Computation:Practice and Experience, 2015.

[91] H. Li, G. Shou, Y. Hu, and Z. Guo, “Mobile edge computing: Progressand challenges,” in 2016 4th IEEE International Conference on MobileCloud Computing, Services, and Engineering (MobileCloud), March2016, pp. 83–84.

[92] Y. Cherdantseva and J. Hilton, “A reference model of information assur-ance amp; security,” in 2013 International Conference on Availability,Reliability and Security, 2013.

[93] J. Somorovsky, M. Heiderich, M. Jensen, J. Schwenk, N. Gruschka,and L. Lo Iacono, “All your clouds are belong to us: security analysisof cloud management interfaces,” in Proceedings of the 3rd ACMworkshop on Cloud computing security workshop. ACM, 2011, pp.3–14.

[94] H. Han, B. Sheng, C. C. Tan, Q. Li, and S. Lu, “A timing-based schemefor rogue ap detection,” IEEE Transactions on Parallel and DistributedSystems, vol. 22, no. 11, pp. 1912–1925, Nov 2011.

[95] D. B. Smetters, D. Balfanz, D. Smetters, P. Stewart, and H. C. Wong,“Talking to strangers: Authentication in ad-hoc wireless networks,”2002.

[96] S. Bouzefrane, A. F. B. Mostefa, F. Houacine, and H. Cagnon,“Cloudlets authentication in nfc-based mobile computing,” in MobileCloud Computing, Services, and Engineering (MobileCloud), 2014 2ndIEEE International Conference on. IEEE, 2014, pp. 267–272.

[97] M. Satyanarayanan, R. Schuster, M. Ebling, G. Fettweis, H. Flinck,K. Joshi, and K. Sabnani, “An open ecosystem for mobile-cloudconvergence,” IEEE Communications Magazine, vol. 53, no. 3, pp.63–70, 2015.

[98] F. Klaedtke, G. O. Karame, R. Bifulco, and H. Cui, “Access control forsdn controllers,” in Proceedings of the third workshop on Hot topicsin software defined networking. ACM, 2014, pp. 219–220.

[99] G. Pek, L. Buttyan, and B. Bencsath, “A survey of security issues inhardware virtualization,” ACM Computing Surveys (CSUR), vol. 45,no. 3, p. 40, 2013.

[100] K. Yang and X. Jia, “Data storage auditing service in cloud computing:challenges, methods and opportunities,” World Wide Web, vol. 15, no. 4,pp. 409–428, 2012.

[101] R. Gennaro, C. Gentry, and B. Parno, “Non-interactive verifiablecomputing: Outsourcing computation to untrusted workers,” in AnnualCryptology Conference. Springer, 2010, pp. 465–482.

[102] C. Wang, N. Cao, K. Ren, and W. Lou, “Enabling secure and efficientranked keyword search over outsourced cloud data,” IEEE Transactionson parallel and distributed systems, vol. 23, no. 8, pp. 1467–1479,2012.

[103] R. Lu, X. Liang, X. Li, X. Lin, and X. Shen, “Eppa: An efficientand privacy-preserving aggregation scheme for secure smart grid com-munications,” IEEE Transactions on Parallel and Distributed Systems,vol. 23, no. 9, pp. 1621–1631, 2012.

[104] S. McLaughlin, P. McDaniel, and W. Aiello, “Protecting consumerprivacy from electric load monitoring,” in Proceedings of the 18th ACMconference on Computer and communications security. ACM, 2011,pp. 87–98.

[105] W. Wei, F. Xu, and Q. Li, “Mobishare: Flexible privacy-preservinglocation sharing in mobile online social networks,” in INFOCOM, 2012Proceedings IEEE. IEEE, 2012, pp. 2616–2620.

[106] J. B. Abdo, T. Bourgeau, J. Demerjian, and H. Chaouchi, “Extendedprivacy in crowdsourced location-based services using mobile cloudcomputing,” Mobile Information Systems, 2016.

[107] M. Sookhak, H. Talebian, E. Ahmed, A. Gani, and M. K. Khan, “Areview on remote data auditing in single cloud server: Taxonomy andopen issues,” Journal of Network and Computer Applications, vol. 43,pp. 121–141, 2014.

[108] J. Shuja, A. Gani, A. Naveed, E. Ahmed, and C.-H. Hsu, “Case ofarm emulation optimization for offloading mechanisms in mobile cloudcomputing,” Future Generation Computer Systems, 2016.

[109] E. Ahmed, A. Gani, M. Sookhak, S. H. Ab Hamid, and F. Xia,“Application optimization in mobile cloud computing: Motivation,taxonomies, and open challenges,” Journal of Network and ComputerApplications, vol. 52, pp. 52–68, 2015.

[110] E. Ahmed, A. Gani, M. K. Khan, R. Buyya, and S. U. Khan, “Seam-less application execution in mobile cloud computing: Motivation,taxonomy, and open challenges,” Journal of Network and ComputerApplications, vol. 52, pp. 154–172, 2015.

[111] E. Ahmed, S. Khan, I. Yaqoob, A. Gani, and F. Salim, “Multi-objectiveoptimization model for seamless application execution in mobile cloudcomputing,” in Information & Communication Technologies (ICICT),2013 5th International Conference on. IEEE, 2013, pp. 1–6.

[112] E. Ahmed, A. Akhunzada, M. Whaiduzzaman, A. Gani, S. H.Ab Hamid, and R. Buyya, “Network-centric performance analysis ofruntime application migration in mobile cloud computing,” SimulationModelling Practice and Theory, vol. 50, pp. 42–56, 2015.

Nasir Abbas received his Master degree in networkand system administration from the Department ofInformatics, University of Oslo in 2017. His main re-search interest is cloud computing for which he hadattended several conferences on cloud technologies.He had worked on OpenStack and Amazon WebServices (AWS) cloud computing to build and main-tain entire IT infrastructures. His research interestsinclude edge computing, 5G networks, mobile cloudcomputing, wireless networks, radio access network(RAN), network functions virtualization and Internet

of Things.


Yan Zhang is Full Professor at the Departmentof Informatics, University of Oslo, Norway. He re-ceived a PhD degree in School of Electrical & Elec-tronics Engineering, Nanyang Technological Univer-sity, Singapore. He is an Associate Technical Editorof IEEE Communications Magazine, an Editor ofIEEE Transactions on Green Communications andNetworking, an Editor of IEEE CommunicationsSurveys & Tutorials, an Editor of IEEE Internet ofThings Journal, and an Associate Editor of IEEEAccess. He serves as chair positions in a number of

conferences, including IEEE GLOBECOM 2017, IEEE VTC-Spring 2017,IEEE PIMRC 2016, IEEE CloudCom 2016, IEEE ICCC 2016, IEEE CCNC2016, IEEE SmartGridComm 2015, and IEEE CloudCom 2015. He servesas TPC member for numerous international conference including IEEEINFOCOM, IEEE ICC, IEEE GLOBECOM, and IEEE WCNC. His currentresearch interests include: next-generation wireless networks leading to 5G,green and secure cyber-physical systems (e.g., smart grid, healthcare, andtransport). He is IEEE VTS (Vehicular Technology Society) DistinguishedLecturer. He is also a senior member of IEEE, IEEE ComSoc, IEEE CS,IEEE PES, and IEEE VT society. He is a Fellow of IET.

Amir Taherkordi received his Ph.D. degree fromthe Department of Informatics, University of Osloin 2011. He is a researcher in the Networks andDistributed Systems group (ND) at the Departmentof informatics, University of Oslo. His research in-terests lie broadly in the area of distributed comput-ing, software engineering, middleware engineering,and embedded systems. His current work is focusedon software architecture, programming abstractions,service distribution, and middleware systems forInternet of Things (IoT). He has experience from

several Norwegian and EU research projects. He was selected as a youngtalented researcher by the Norwegian Research Council in 2017 to work on anovel IoT service computing model for future Fog-Cloud computing systems.

Tor Skeie received the MS and PhD degrees in com-puter science from the University of Oslo in 1993and 1998, respectively. He is a professor at the Sim-ula Research Laboratory and the University of Oslo.His work is mainly focused on scalability, effectiverouting, fault tolerance, and quality of service inswitched network topologies. He is also a researcherin the Industrial Ethernet area. The key topics herehave been the road to deterministic Ethernet endto end and how precise time synchronization canbe achieved across switched Ethernet. He has also

contributed to wireless networking, hereunder quality of service in WLAN,and cognitive radio.

ieee internet of things journal, vol. aa, no. bb, mmmm...

Documents