By: Faizan Agha (DESEM)Supervisor: Dr Stephen Brown

Traditional Communication Networks◦ With infrastructure (e.g. Access Point, Ethernet..)◦ Without infrastructure (e.g. Mobile Ad hoc Networks

- MANETs)Infrastructure-based Network [1]

Ad hoc Network (MANET)

Opportunistic Networks:◦ Class of Delay Tolerant Networks (DTNs)◦ End-to-end path or route does not exist ◦ Suitable for communications which can afford longer delays◦ Make use of Users’ mobility and their mutual encounters◦ ‘Store-carry-and-Forward’ Architecture◦ Energy constraints for mobile nodes

Figure 1 An example of DTN: Helicopter being a Gateway Node for two discrete internets

Figure 2 Every node (or host or router) relays opportunistically. Notice a variety of contributors

Applications in remote area and emergency situations (e.g. Recent earthquake in Nepal)Most of the past simulations assume that nodes always behave altruisticIntroduce a configurable plug-in for an open source simulator, to help leverage further implementationsSupport The ONE community for Research and Development w.r.t OppNets

Impact of Rogue Nodes on reliability of an opportunistic network:Metric Average Latency (by simulating a portion of analytical modelling and numerical results in [4])

[1]

Performance Modeling for Two-hop Relay with Node Selfishness in Delay Tolerant Networks [4]◦ Modeling by two dimensional Continuous Time

Markov Chain-based (CTMC) mathematical framework◦ Delay is quantified as:

Rogue node Behavior

Probability of refusing to accept and store data meant

for another node

P nc

Refusing to relay data

P nf

The Impact of Node Selfishness on Multicasting in Delay Tolerant Networks [6]◦ Theoretical and Simulation.◦ Epidemic and Two-hop relayingIRONMAN: Using Social Networks to Add Incentives and Reputation to Opportunistic Networks [7]◦ A possible approach to increase altruism and

discourage malicious nodes

•Reverse Engineering of ONE Simulator Workflows and APIs

•ONE Online community Membership

•Sample runs with default configuration

ONE Simulator Investigation

•Platform Setup on Windows/Eclipse

•Customization of Core/Routing APIs for Rogue Plug-in

Rogue Behavior Implementation •Configuration of

parameters w.r.tanalytical model

•Running Experiments with carying Rogree DegreeExperiments Design and Execution

•Data Set Analysis•Deductions and

Conclusions•Future Work

Results and Deductions

Opportunistic Network Environment (ONE*) is developed in Java with hundreds of APIs and Post ProcessingIntroducing rogue behavior in ‘Simulation Engine’ and ‘Routing Logic’ by developing plug-in for: Rogue behavior, Rogue Degree.

*ONE Simulator is open source and Supported by Nokia Research Center (Finland) [3]

MessageRouter

PassiveRouter

ActiveRouter

Epidemic DirectDelivery

SnW ...

Figure 9 Hierarchy of Message Routers in ONE

Figure 8 Rogue Plug-in impacted packages (Highlighted blue)

• Performing a series of experiments with 100 seeds to:Verify correctness of changes incorporated for configurable rogue behaviorCompare analytical and simulation results produced in related work [4]

ExperimentNo.

RogueDegree

Rogue Flag

1 0 True2 20 True3 40 True4 60 True5 80 True6 90 True7 100 True

ScenarioEnd Time(s)

HostGroups

RogueBehavior

TransmitterRange(m)

TTL(s)

Seeds WalkingSpeed (m/s)

43200 1 True 50 43200 100 0.5 to 2.5

Figure 17 Number of experiment with varying rogue degreeFigure 18 Some important input parameters

1 2 3 4 5 6 7

Degree 0 20 40 60 80 90 100

latency 5904.78 5900.69 5873.77 5895.52 5860.3 5917.01 11039.7

0

2000

4000

6000

8000

10000

12000

Aver

age

Late

ncy

in S

econ

ds

Percentage of Rogue Node nodes with increasing degree

Average Latency against increasing Rogue Degree: Epidemic Router

Figure 26 Iteration 2 of Average Latency versus Rogue Degree. Each experiment with the respective degree was run with 100 seed values

Figure 25 Our evaluation was targeted against the black dotted line representing Epidemic Router with Pnc = 1 and N = 50. As

discussed in Section 5.3.

CTMC distribution based Rogue Nodes were not implemented and simulatedOnly Probability value 0 – 1 could be verified.Comparison with Two Hop Relay could not be done due to limited scopeValue of lemda (contact duration) is not configurable in the simulator. We made use of secondary parameters to approximate that.

In smaller geographical areas, Mobile Nodes forwarding messages under Epidemic Routing are inherently resilient to rogue behavior up until 90% of degree.Simulated results showed correlation with the analytical modeling till almost 55% of nodes are rogue.Contribution for ONE repository: A configurable plug-in for plenty of future implementations:◦ Comparison with Two Hop Relay Routing Algorithm or any other

comparable algorithms like SnW, PROPHET or DD routers.◦ Simulation with real world traces will provide more realistic data

set. Time taking process.◦ Automation for varying rogue degree and graph generation◦ Introduction of incentives or penalization schemes to encourage

or discourage rogue behavior in an OppNet.

1. http://www.ece.ncsu.edu/netwis/Fundamental_Properties.php2. http://www.slideshare.net/noorin/opportunisticnetworking3. http://www.netlab.tkk.fi/tutkimus/dtn/theone/4. M. Karaliopoulos, "Assessing the Vulnerability of DTN Data

Relaying Schemes to Node Selfishness," 2009. 5. The Effects of Node Cooperation Level on Routing Performance

in Delay Tolerant Networks Resta, G. ; IIT-CNR, Pisa, Italy ; Santi, P.

6. The Impact of Node Selfishness on Multicasting in Delay Tolerant Networks Yong Li ; Dept. of Electron. Eng., TsinghuaUniv., Beijing, China ; Guolong Su ; Wu, D.O. ; Depeng Jinmore authors

7. IRONMAN: Using Social Networks to Add Incentives and Reputation to Opportunistic Networks Bigwood, G. ; Sch. of Comput. Sci., Univ. of St. Andrews, St. Andrews, UK ; Henderson, T.

Thanks

On every invocation of the message receipt:Check if the rogue behavior is enabled in Input Configuration File,

If NO thenPut the message in the buffer and receive the message as per the protocol

If YES thenIf the Router is willing to relay or store as per Pseudorandom Generator

if YES then continue receiving normallyIf NO then check if the message was intended for the router itself

if YES then put the message in the incoming bufferIf NO then deny the message as rogue with probability P (0-100)

On every instance of a contributing Node in an Oppnet:Generate a random number (0-100) via the global generator;Compare the value with the input Rogue Degree

If the random number is > Rogue DegreeNode will relay the message as being altruistic

elseNode will deny the message as ROGUE_DENIED

On every set up of the scenario for the simulation run:If the Rogue Behavior is disabled

Perform simulation with altruistic nodeselse

Keep all the simulation settings and parameters sameCreate all the nodes being rogue, associating each with a Rogue Degree from the input file

1. Routing implementation for Rogue Nodes

2. Node Willingness decision

Figure 15 Snapshot of ONE simulator running in GUI mode. The green circles represent the range of the respective node, denoted by p<n> where n is the sequence number. Each scenario is listed as the Title, CS‐645‐0‐1 in this case.

Figure 20 For Rogue Degree 20 the Message Stats Reports (All the metrics explained in Section 7.7.3)