1. Sogang University ICC Lab Using Game Theory to Analyze Wireless Ad Hoc networks

2. Sogang University ICC Lab. 2 Outline Ad-Hoc network Game theoryAd-Hoc + Game theory Social optimal Medium access layer Network layer Transport layer Physical layer Layered approach Future work

3. Sogang University ICC Lab. 3 Ad Hoc networks What are ad hoc networks  Multi-hop communication  Reduced need for any infrastructure  Dynamic topology  Distributed, interactive stations  Ease of deployment  Potentially more robust to attack Application of ad hoc networks  Military application  Disaster management  Impromptu communication between people

4. Sogang University ICC Lab. 4 Game Theory Game theory – a branch of mathematics used extensively in economics The study of mathematical models of conflict and cooperation between intelligent rational decision makers- (1991) Basic component: Game – A mathematical representation of an interactive decision situation Important concepts  Conflict and cooperation  Intelligent rational decision makers

5. Sogang University ICC Lab. 5 Basic component Strategic game – 3 basic components  A set of 2 or more players (N = {1,2,….n})  A set of actions for each player ( )  Utility function for every player ( ) Nash equilibrium  An action vector is a Nash equilibrium if and  An action vector from which no player can benefit by deviating unilaterally A B Confess Not confess ConfessNot confess 5,50,15 15,01,1 NE Prisoner’s dilemma

6. Sogang University ICC Lab. 6 Why game theory? De-centralized nature of nodes  Independently adapting its operation based on perceived or measures statistics Interactive decision makers  Decision taken by one node affects and influences the other nodes Available Adaptations MANET Component Game Component Action Set Nodes in NetworkPlayer Set Adaptation Algorithm Decision Update Algorithm Valuation Function (Preference Relations) Utility Function Learning Process

7. Sogang University ICC Lab. 7 Steps in application of game theory Develop a game theoretic model  Solution of game’s Nash equilibrium yields information about the steady state and convergence of the network Does a steady state exist?  Uniqueness of Nash equilibrium Is it optimal? Do nodes converge to it? Is it stable? Does the steady state scale?

8. Sogang University ICC Lab. 8 Optimal equilibrium inducing mechanisms Credit exchange  Virtual currency ▪ Difficult to implement  Reputation ▪ Appropriate for denial of service attacks Other schemes  Generous Tit-for-tat ▪ Node mimics the action of its peers ▪ Slightly generous  Watchdog mechanism ▪ Specific to prevent malicious/selfish behavior in routing Presence of centralized referee  Not a player but an overseer  Not a typical game theoretic scenario

9. Sogang University ICC Lab. 9 Physical and Medium access layers Power control  Adjust transmit power levels  Objective: To achieve a target signal-to-interference-to-noise ratio Medium access  Set the probability of packet transmission  Objective: To maximize individual throughput

10. Sogang University ICC Lab. 10 Network layer (Research issues) Previous work restricted to analyzing selfish node behavior while forwarding of packets  Nodes decide on the proportion of packets/sessions to act as a relay  Energy is the main constraint  “Selfishness is the only strategy that can naturally arise in a single stage” Use of external incentive mechanisms to induce socially optimal equilibrium Shortcomings  Do not consider true ad hoc scenarios where nodes can experience inherent trade-offs  Do not consider mobility and influence on entire network  Restrict the model to relaying packets

11. Sogang University ICC Lab. 11 Network layer (Current research) Node participation  Switch interfaces to a sleep state  Affects network operations ▪ Network partition ▪ Network congestion  Individual losses ▪ Loss of information for an ongoing session ▪ Overhead involved in discovering location of other nodes on waking up ▪ Extra flow of route queries due to frequent topology changes

12. Sogang University ICC Lab. 12 Network layer (Other issues) Malicious node behavior degrades performance of dynamic source routing protocol Classic routing  Nodes decide on the amount of data to be sourced on shared paths to minimize the cost involved Use of game theory – infant stage

13. Sogang University ICC Lab. 13 Transport layer Analyze congestion control algorithms for selfish nodes  Objective: Determine the optimal congestion window additive increase and multiplicative decrease parameters  Current efforts restricted to traditional TCP congestion control algorithms for wired networks Ad hoc networks  Incorporate the characteristics of the wireless medium in the congestion control game

14. Sogang University ICC Lab. 14 Summary Game theory offers a promising set of tools to analytically model ad hoc networks Game theory can be used  Analysis of ad hoc networks  Design of incentive mechanisms Past research concentrated on wired/cellular networks Design of robust protocols to deal with selfish behavior Not completely realistic : Difficult to obtain social optimum in a distributed environment of rational entities

15. Sogang University ICC Lab. 15 Future Work Currently developing a model for node participation in an ad hoc network Analyze the model using game theoretic techniques and determine the optimal in the ad hoc network Apply the node participation model to a well known routing protocol and study the effect of varying level of node participation Incorporate mobility in the game theoretic model