Reputation Methods for Manets

1
Reputation Methods for Manets

• Jiangyi Hu

2
Outline

• Two reputation systems
• Issues with reputation method
• Local aware reputation system
• Calculate and update reputation values
• Detection of misbehavior

3
CORE (Michiardi Molva, 2002)------
Collaborative Reputation Mechanism

• Three types of reputation
• Subject reputation, locally calculated
• Indirect reputation, accepted from other nodes
• Functional reputation, related to certain
  function
• Three ways to update reputation
• Local calculation of subjective reputation
• Global distribution of reputation
• If no interaction, the reputation gradually
  decreases

4
CONFIDANT (Buchegger Boudec, 2002)
----Cooperation of Nodes Fairness In Dynamic
Ad-hoc NeTworks
5
Issues with reputation method

• Intentional vs. accidental misbehavior
• Inconsistent reputation value
• Trust vs. reputation
• Trust how honest is a node (trustworthy / not
  trustworthy)
• Reputation how well a node behaves (regular /
  misbehaved)
• Direct vs. indirect reputation
• Global vs. local reputation

6
Direct indirect reputation

• A principal may consider another principal
• trustworthy (1)
• not trustworthy (-1)
• unsure of the trustworthiness (0)

7
Direct indirect reputation

• Trust may not be symmetric.
• Trust is usually transitive.
• A node A can know something about another C from
  the indirect reputation message if and only if it
  knows the message provider B is trustful.

1/-1
1/-1
1
1/-1
-1(0)
0/0
A
B
C
8
Direct indirect reputation

• A node A can know something about a reputation
  message provider B from the indirect reputation
  message if and only if what B says about C is
  contrast from what A knows C is.

1
-1
1/-1
0/-1
1/-1
-1/0
A
B
C
9
Global vs. local reputation

• Global
• Every node knows reputation of every other node
  in the network
• Large storage at each node, more network traffic
• Good node here does not mean good node elsewhere
• Nodes does not care about reputation of distant
  nodes
• Local
• Nodes aware of reputation of its k-hop
  neighbors

10
Local aware reputation system

• Number of malicious node lt K
• Blacklist is signed by K neighbors
• Blacklist broadcast in m-hop neighbors

A
E
D
M
B
C
11
Model notation

• Destination acknowledgement ack
• When no acknowledgement received trace
• Neighbors participate in tracing faults
• Reputation value R Trust levels
• Rmin lt RN lt Rmax
• Rmin lt Rt lt RN lt Rmax N is trustworthy
• Rmin lt RN lt Ru lt Rmax N is not
  trustworthy
• Rmin lt Ru lt RN lt Rt lt Rmax unsure of Ns
  trustworthiness
• N(X) 1-hop neighbors of node X
• RM Reputation value of node M

12
Detect misbehavior update reputation

• Case 1 M does not forward message
• X RM RM a X
  in N(M)nN(N)

M
N
13
Detect misbehavior update reputation

• Case 2 M does not forward message nor trace
• X RM RM b X
  in N(M)nN(N)
• b gt a

M
N
14
Detect misbehavior update reputation

• Case 3 M does not forward message, but forwards
  trace, M and M collude and cheat
• X RM RM c
  X in N(M)nN(M)
• RM
  RM c X in N(M)nN(M)
• RM
  RM a X in N(M)-N(M)
• 
  c gt b gt a

M
M
N
15
Conclusion

• Reputation based on direct observation
• Reputation message is always trustworthy
• Less problem with inconsistent reputation value