A Reputation Based Scheme for Stimulating Cooperation

1
A Reputation Based Scheme for Stimulating
Cooperation

• Xin Wang
• Assistant Professor
• Department of Electrical Engineering
• Stony Brook University
• (Joint work with Aruna Balasubramanian and Joy
  Ghosh)

2
Outline

• Problem definition
• Related Research
• Our Solution Reputation based solution for
  stimulating cooperation
• Reputation System
• Monitoring System
• Cooperation System
• Security System
• Conclusions
• References

3
Problem definition Ad hoc networks
Self organized network
Lack of central authority to coordinate routing
Wireless links
B
E
C
Communication between A and B using C Multihop
D
A
C routes packets for A and B
Stimulating nodes to cooperate is important to
ensure optimum network utilization
4
Problem definition Reasons for Non cooperation
Node
Malicious
Genuine
Selfish
Resource Constraint
Irrationally malicious
Rational malicious
5
Related Research

• Using Incentive Management schemes Buttyan
  2000, Zhing 2003
• Give incentive to nodes that have a good credit
• by providing routing
• Nodes in unfavorable location may not get chance
  to transmit packets
• Difficult to determine the amount to
  credit/debit,
• Difficult to ensure security of the credit system

• Punishment based schemes Buchegger 2002, Marti
  2000,
• Punish misbehaving nodes
• Cooperating nodes have no incentive to continue
  cooperation
• Congestion and other reasons may lead to wrong
  punishment
• A misbehaving node can move out from the current
  neighborhood to a new neighborhood, to avoid
  punishment

• Game theoretical approaches 7
• Useful in analyzing the cooperation solution

6
Our solution Both rewards well behaved nodes and
punishes non cooperation
COMPONENTS
MONITOR SYSTEM Monitors neighbors nodes for
packet dropping and forwarding
Distribute reputation to neighbors
Reputation reports from neighbors
COOPERATION SYSTEM Punish nodes with low
reputation Give incentive to nodes with high
reputation Use reputation of node to
ensure reliability
REPUTATION SYSTEM Calculates reputation
SECURITY SYSTEM Ensures the security of all the
components
7
Reputation System

• Reputation Goodness of a node as perceived by
  its neighbors
• New node has neutral reputation
• If Reputation lt Threshold, node is punished
• Reputation increased for good behavior at the
  rate of a and decreased for bad behavior at the
  rate of ß
• High a Node builds up reputation faster and
  misbehaves for prolonged time
• Low a Not enough incentive
• High ß Genuine node that drops packets due to
  network failure will be punished
• Low ß Takes a longer time for misbehaving nodes
  to be punished

8
Reputation System (Contd)

• To solve this, in our solution,
• a and ß are chosen carefully according to the
  network characteristics
• The reputation is not reduced or increased
  linearly, but as a function of the nodes current
  reputation and the number of packets
  dropped/forwarded
• Smaller the reputation, smaller is the number of
  packets that are dropped to reduce reputation
• Larger the reputation, more is the number of
  packets that need to be forwarded to increase the
  reputation
• Reputation value is changed cumulatively, at
  regular intervals, and not every time a packet is
  dropped or forwarded

9
Monitoring System

• Neighbor monitors nodes to check if node forwards
  packets

A transmits packet (sent by B) to C
B listens to this transmission
B
A
If B does not hear its packet being transmitted
for a while, it assumes that the A has dropped
it As reputation is re-calculated
C
10
Monitoring System (Contd)

• Reputation calculation based on own observation
  alone may not be sufficient

• Every node distributes the reputation it observes
  to all neighbor nodes, to ensure that all nodes
  have a consistent view about each other

11
Monitoring System (Contd)

• Every node calculates reputation as a weighted
  mean of its own observation and the neighbor
  reports
• Weights given to a neighbor report is
  proportional to how much the neighbor is trusted

A
C
B
Bs reputation
D
12
Monitoring System (Contd)

• Common problems with monitoring systems
• Distribution of false reputation reports by
  malicious neighbors
• In our solution, false reputation reports are
  given less weight and thus their effect will not
  be significant
• Incorrect monitoring, when packets are dropped
  due to congestion or collision
• We implement a mechanism to identify congestion
• Incorrect penalty due to incorrect monitoring is
  reduced considerably due to our tolerance scheme

13
Cooperation System Penalty

• If the neighbor node has reputation lower than a
  threshold
• Do not forward any packet for this neighbor
• Re route packets, if the next hop is the
  misbehaving neighbor
• Common problem with cooperation systems is the
  inability of a repentant node to rejoin the
  network
• We provide alternate protocols for repentant
  nodes to rejoin the network
• Idle protocol Node finishes penalty time and
  joins the network with neutral reputation
• Redeem protocol Node participates in forwarding
  packets, and can start sending its own packets
  when its reputation increases to the neutral
  reputation

14
Cooperation System Incentive

• Intermediate nodes prioritize packets based on
  the reputation of source and destination

Source sends first packet with the certificate of
the source and destination
Packet n
Intermediate nodes stores certificate
Packet 2
Source
Destination
Intermediate nodes prioritize subsequent packets
based on the reputation of the source/destination
Certificate certifies the reputation of a node by
a trust mechanism
15
Security System

• Obtaining certificates in order to get incentive
  and ensuring that malicious nodes do not obtain
  the certificate
• Certification should be decentralized
• Nodes should be able to carry their incentive
  when they leave their neighborhood

• Nodes provide certificates to neighbors
• Using threshold cryptography
• Divide the certification authoritys private key
  among all nodes
• Some k nodes need to combine their keys to create
  the certificate
• No combination of less than k nodes can create
  the certificate

• Nodes in the new neighborhood turn a flag on,
  when any new node sends a packet
• Nodes in the old neighborhood, on seeing the flag
• send the real reputation of the node in the route
  reply

• Prevent non cooperative node from leaving the
  neighborhood to a new neighborhood where their
  misbehavior has not been reported

16
Results

• Simulated the cooperation solution in GloMoSim
  (Network Simulator)
• Solution simulated in the routing layer, built
  over Dynamic Source Routing Protocol (DSR)
• Nodes are randomly assigned as being
  non-cooperative, and these nodes drop packets
• Some parameters
• Number of nodes 50
• Network size 1000 m X 1000 m
• Simulation time 40 min
• Radio range 250 m

17
Results (Contd)
The decrease in throughput (of cooperating nodes)
is significantly lower when using our cooperation
scheme, even when the percentage of non
cooperating nodes is as high as 30
18
Results (Contd)
The end-to-end delay of cooperating nodes is
significantly lower, when using our cooperation
scheme, due to incentives provided to well
behaved nodes
19
Results (Contd)
Since packets of nodes with low reputation are
dropped, the throughput of packets sent by non
cooperating nodes is very low, and this is a
punishment
20
Results (Contd)
Non cooperating nodes (even those whose
reputation gt threshold) are provided with lower
service compared to cooperating nodes and their
packets take longer to reach
21
Conclusions

• In this work, we provide a novel reputation
  scheme that rewards well behaved nodes and
  punishes non cooperative nodes, in an effort to
  stimulate cooperation
• Reputation System
• Monitoring System
• Cooperation System
• Security System
• Simulation results indicate that the cooperation
  solution significantly improves performance of
  well behaved nodes, in terms of throughput and
  end-to-end delay

22
References

• L. Buttyan and J.-P. Hubaux. Enforcing Service
  Availability in Mobile Ad-Hoc WANs. In
  Proceedings of the IEEE/ACM Workshop on Mobile Ad
  Hoc Networking and Computing (MobiHOC) , Boston,
  August 2000.
• Hubaux, J., Gross, T., Le Boudec, J., Vetterli,
  M. Towards self-organized mobile ad hoc networks
  The Terminodes project. IEEE Communications
  Magazine, (January 2001).
• S. Zhing, J. Chen and Y.R. Yang, SPRITE A
  Simple, Cheat-Proof Credit-based System for
  Mobile Ad hoc Networks, in Proceedings of IEEE
  INFOCOM 03, San Fransesco, CA, April 2003.
• S. Marti, T. J. Giuli, K. Lai, and M. Baker,
  "Mitigating routing misbehavior in mobile ad hoc
  networks," in Sixth annual ACM/IEEE International
  Conference on Mobile Computing and Networking,
  2000, pp. 255--265.
• S. Buchegger and J. Le Boudec. Nodes Bearing
  Grudges Towards Routing Security, Fairness, and
  Robustness in Mobile Ad Hoc Networks. In
  Proceedings of the Tenth Euromicro Workshop on
  Parallel, 403 -- 410, Canary Islands, Spain,
  January 2002. IEEE Computer Society.
• P.Michiardi and R.Molva, CORE A Collaborative
  Reputation Mechanism to Enforce Node Cooperation
  in Mobile Ad hoc Networks, in Proceedings of the
  IFIP TC6/TC11 Sixth Joint Working Conference on
  Communications and Multimedia Security. Kluwer,
  B.V., 2002, pp. 107 - 121
• V.Srinivasa, P.Nuggehalli and C.Chiasserini,
  Cooperation in Wireless Ad hoc Networks in
  Proceedings of IEEE INFOCOM 03, San Fransesco,
  CA, April 2003.

23
Problem definition Non cooperation

• Non cooperation nodes in the network refuse to
  cooperate in providing network services
• Significant amount of packet dropping will reduce
  the network throughput
• Ad Hoc Environment (lack of central control,
  existence of implicit trust) increases challenges
  in isolating non-cooperative nodes
• Stimulating nodes to cooperate is important to
  ensure optimum network utilization

24
Results Metrics used

• Throughput of cooperating nodes () Number of
  packets sent by cooperating nodes / Number of
  packets received x 100
• Throughput of non cooperating nodes () Number
  of packets sent by non cooperating nodes / Number
  of packets received x 100
• Average End-to-End delay of cooperating nodes
  (seconds) Total delay for packets sent by
  cooperating nodes to reach the destination /
  Total number of packets
• Average End-to-End delay of cooperating nodes
  (seconds) Total delay for packets sent by non
  cooperating nodes to reach the destination /
  Total number of packets

25
Related Research Limitations

• Incentive Management Schemes
• Nodes in an unfavorable location, that cannot
  participate in routing, will not be allowed to
  route its own packets
• Difficult to determine the amount to
  credit/debit, using a decentralized algorithm
• More difficult to ensure the security of the
  credit system

• Punishment Based schemes
• Cooperating nodes have no incentive to continue
  cooperation
• Genuine nodes that drop packets due to network
  congestion and other reasons may be wrongly
  punished
• A misbehaving node can move out from the current
  neighborhood to a new neighborhood, to avoid
  punishment

26
Monitoring System (Contd)

• Every node calculates reputation as a weighted
  mean of its own observation and the neighbor
  reports
• Weights given to a neighbor report is
  proportional to how much the neighbor is trusted

A receives reputation report of B, from C and D
A has three reputation of B (including its own) ,
RepAB, RepCB and RepDB
A
Weight given to reputation report of C by A
Reputation of C with A, RepAC / (RepAB RepCB
RepDB) Weight given by A to its own reputation
Ut (Maximum reputation)
C
B
Bs reputation
D
A calculates the reputation of B using the
reputation reports and the weights given to them