Office of Naval Research Project Node Participation In MANETs

1
Office of Naval Research Project Node
Participation In MANETs

• Dr Luiz DaSilva
• Vivek Srivastava
• December 8, 2003

2
Agenda
3
Objectives

• Better understanding of node behavior and
  participation in MANETs
• Design of incentive mechanisms to induce selfish
  nodes towards a desirable equilibrium

4
Literature Review

• Limited amount of research
• Recent work provides a basis for further
  development of game theoretic models for MANETs
• Potential challenges
• De-centralized environment
• Node mobility
• Dynamic link establishment

5
Previous work (1)

• Node participation (Forwarding of packets)
• V. Srinivasan et al., Cooperation in Wireless Ad
  Hoc Networks, Proceedings of IEEE INFOCOM,
  vol.2, April 2003, pp. 808-817.
• A. Urpi, M. Bonuccelli and S. Giordano, Modeling
  cooperation in mobile ad hoc networks a formal
  description of selfishness, Proceedings of the
  Workshop on Modeling and Optimization in Mobile
  and Wireless Ad Hoc networks, March 2003.
• Security
• P. Michiardi and R. Molva, Game theoretic
  analysis of security in mobile ad hoc networks,
  Institute Eurecom, Sophia-Antipolis, France,
  Report no.RR-02-070, April 2002.

6
Previous Work (2)

• Management of Sensor networks
• L. Johansson, N. Xiong and H. Christensen, A
  Game Theoretic Model for Management of Mobile
  Sensors, Proceedings of the Sixth conference on
  Information Fusion, July 2003, pp. 583-591.
• Incentive mechanisms
• S. Buchegger and J.Y. Le Boudec, Performance
  Analysis of the CONFIDANT Protocol Cooperation
  Of Nodes Fairness In Dyanamic Ad-Hoc NeTworks,
  Proceedings of ACM MobiHoc, June 2002.
• L. Buttyan and J. P. Hubaux, Nuglets A Virtual
  Currency to Stimulate Cooperation in
  Self-Organized Mobile Ad-Hoc Networks, Swiss
  Federal Institute of Technology, Lausanne,
  Switzerland, Report no. DSC /2001/001, January
  2001.

7
Potential Issues in MANETs (1)

• Node participation
• Different from denying forwarding requests,
  although end result may be similar
• A network designer may not want all nodes to
  participate all the time to increase the
  longevity of the network

Network congestion
Network Partition
8
Potential Issues in MANETs (2)

• Routing in MANETs
• Robustness of routing protocols
• Energy constrained nodes deny forwarding/routing
  requests
• Results in additional route queries and loss in
  throughput
• Investigate techniques to improve the robustness
  of standard protocols (DSR,OLSR, AODV) for such
  nodes
• Security
• Investigate Denial of Service (DoS) attacks in an
  ad hoc network in the presence of malicious nodes
• Severe implications in the form of misguided
  routing
• Design mechanisms to punish such behavior

9
Other issues

• Congestion Control
• Typically, wired networking games evaluate the
  TCP congestion control parameters at steady state
  for selfish node behavior
• Link failure could be misinterpreted as
  congestion in an ad hoc network
• Classic competitive routing
• Determine the optimal flow configuration for
  nodes sharing links between a source and
  destination in ad hoc network
• Application Improving the performance of
  multi-path routing for ad hoc networks under
  competitive behavior

10
Potential Applicability

• Energy constrained nodes (e.g. sensors)
• Malicious nodes that cause jamming, interception
  and other intrusive actions

11
Voluntary resource sharing

• Definition
• All nodes voluntarily help each other
• to achieve a network-wide goal
• Sharing of resources such as storage, bandwidth,
  processing capabilities
• Examples ad hoc networks, sensor networks, grid
  computing, peer-to-peer networks
• Objective
• Analyze node behavior under benefits/threats
  perceived by nodes in sharing
• Extend its applicability to study node
  participation in ad hoc networks

12
Voluntary resource sharing

• Single shot game (played once)
• Homogeneous nodes
• Selfish nodes perceive a cost in sharing
• Heterogeneous nodes
• Some nodes gain by sharing or are indifferent to
  sharing
• Presence of a rogue node
• Rogue node malicious intent
• Repeated games
• Intuitively, if all nodes were selfish the Nash
  equilibrium would be for the nodes not to share
  their resources
• If such were the observed behavior voluntary
  resource sharing networks would not exist

13
Model parameters
14
Model parameters (contd)
15
Heterogeneous users

• For any user
• For any user j
• The equilibrium is for users that perceive a cost
  in sharing to not share their files and those
  that benefit or do not perceive a cost in
  sharing, to share.

16
Rogue nodes

• A node has incomplete information regarding the
  type of the other node is it a regular node (A
  node that gains by sharing under perfect
  information) or a rogue node?
• Model it as a game of incomplete information
  (Bayesian game)
• Define additional parameter to indicate the
  type of player
• tj 0 Regular node tj 1 Rogue node

17
Rogue nodes

• Each node sets its belief about the type of the
  other node
• Define as the conditional probability
  assigned by node j to the type of the other node
  if node j is of type tj
• The Bayesian game can be expressed as
• For a regular node j (P signifies a penalty and
  R signifies a reward)
• Let be the probability assigned by node j
  that there is a rogue node in the network

18
Bayesian Game Result

• We assume that it is a dominant strategy for
  rogue nodes to share.
• The expected utilities for the regular node are
• The regular node will benefit by sharing its
  resources only when
• As expected, if the node will not benefit
  by sharing.

19
Repeated game

• Nash equilibrium for homogeneous nodes is
  undesirable
• Finite repetition of the game may result in a
  desirable equilibrium
• Consider K repetitions, K is a random variable
  with geometric distribution (memoryless)
• Strategy of nodes Share as long as other nodes
  share, do not share if any other node deviated in
  previous round

20
Interpretation

• If , it is an equilibrium to share
• If , an altruistic equilibrium
  depends on benefit derived from others
  resources, cost of sharing ones own and time
  horizon

21
Repeated game result

• Revisit
• If no node deviates, expected pay-off from that
  stage forward is
• If any node deviates, expected pay-off from that
  stage forward is
• Nodes will share (equilibrium) as long as,

22
Applicability to MANETs

• Extend results to heterogeneity observed in a
  MANET
• Cluster management
• Cluster heads could benefit by sharing their
  resources as it keeps the traffic in the cluster
  to a minimum
• Nodes that track your behavior could be kept at
  bay by estimating the probability of their
  presence
• If mobility is low, a node must rely on same set
  of neighbors for routing or other services for
  extended periods corresponding to
• (repeated game)
• Correlation with node participation - nodes share
  their resources when they decide to participate

23
Node Participation

• Nodes experience a trade-off by controlling the
  amount of participation in the ad hoc network
• Benefits
• Increased lifetime of nodes (inversely
  proportional)
• Increase in throughput by participating (directly
  proportional)
• 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

24
Node Participation

• Objective Characterize node behavior and
  determine the optimal time a node stays awake in
  an ad hoc network
• Study the manifestations on standard routing
  protocols, service provisioning and cluster
  management

25
Model parameters
26
Model parameters

• Utility function

27
Future Work

• Immediate goals
• Analyze the trade-offs involved in node
  participation to determine if nodes would
  actively participate
• Characteristics of the operating point
• Will node behavior converge to the desirable
  operating point?
• Is the operating point maintained even if a few
  nodes reduce their level of participation from
  it?
• Optimality check - does the network benefit
  (overall throughput vs longevity)?

28
Future Work

• Long term goals
• Suggest incentive mechanisms to induce nodes to
  participate and achieve desirable equilibrium
• Formalize the node participation game and study
  its manifestations in routing and service
  discovery
• Consider the effect of mobile nodes on the game
• Validate results with simulation