Robust Incentives via Multilevel Titfortat

1
Robust Incentives via Multi-level Tit-for-tat
IPTPS, Feb. 2006

• Qiao Lian, Zheng Zhang (MSRA)
• Yu Peng, Mao Yang, Yafei Dai, Xiaoming Li (PKU)

2
P2P file-sharing needs incentives to work

• genuine incentives must collaborate/share to
  benefit
• E.g. block exchange in BT
• Problems only works within a large session

• Nearly 80 sessions contain 2 peers only, i.e.
  there is only one downloader
• No one else to collaborate with!

3
A simple breakdown of the spectrum

• Artificial incentive
• Produce/record evidence of collaboration for
  future reference

Brittle to collusion and other problems
Incentives
Shared history
Private history
Artificial incentives
genuine incentives
subjective
Non-subjective
Absolute contribution (e.g. Maze)
The sum of contribution from the perspective of
other peers, weighted by their reputation (e.g.
EigenTrust)
4
Talk organization

• The Maze p2p file sharing system
• Existing collusion behaviors
• Why simple algorithms do not work
• EigenTrust and Tit-for-Tat
• Multi-trust algorithm
• Evaluation
• Summary and Related work
• Conclusion

5
Mazearchitecture

• A sends query
• server responses with file / replica info
• A sends download requests
• B and C response with file data
• B and C upload traffic log

centralize maintained index / membership
user cloud
Our work starts from these logs
C
A
B
6
Vital statistics

• Popular
• Population 1.4 million registered accounts
  30,000 online users
• More than 200 million files
• More than 13TB (!) transfer everyday
• Completely developed, operated and deployed by an
  academic team
• Logs added since the collaboration w/ MSRA in
  2004
• Enable detailed study at all angles

7
MazeIncentive Policies

• New users points 4096
• Point change
• Uploads 1.5 points per/MB
• Downloads at most -1.0 point/MB
• Gives user more motivation to contribute
• Benefit of high point
• Climbing ladder ? social status
• Service differentiation
• Order download requests by T Now 3log(Point)
• Users with P 200Kb/s
• Available in Maze5.0.3 extensively discussed in
  Maze forum before implemented

8
Talk organization

• The Maze p2p file sharing system
• Existing collusion behaviors
• Why simple algorithms do not work
• EigenTrust and Tit-for-Tat
• Multi-trust algorithm
• Evaluation
• Summary and Related work
• Conclusion

9
What is collusion

• Definition (Webster dictionary)
• secret agreement or cooperation especially for an
  illegal or deceitful purpose
• And in the Maze context
• Multiple peers collude to defeat the incentive
  system
• What makes the study hard
• Even with all the traffic logs, we will never
  know for sure
• But we can identify suspicious colluding patterns
• See our technical report for more details

10
the collusion workingset
221,000 pairs whose duplication degree 1
the top 100 pairs with most redundant traffic

• Repeat traffic detector
• Hint colluders are lazy
• for peer pair link duplication degree total
  traffic / unique data

11
A closer look
Ted 3.8TB
Sam 47GB
Ingrid 78GB
Mary 73GB
Star-shape collusion (spam account) colluding
whitewashing account
(Fred, Gary)
(Olga, Pam)
Pair-wise collusion
(David, Alice, Quincy) e.g. Alice uploads MSDN
DVD image (3GB) for 29 times
(Harry, Cindy)
12
Talk organization

• The Maze p2p file sharing system
• Existing collusion behaviors
• Why simple algorithms do not work
• EigenTrust and Tit-for-Tat
• Multi-trust algorithm
• Evaluation
• Summary and Related work
• Conclusion

13
What about EigenTrust?

• EigenTrust clone of PageRank
• Basic idea
• Consider recommenders reputation
• Trust matrix M
• mi,j trust of peer i to peer j (e.g download
  quantity)
• normalize each row of M
• EigenTrust vector
• The left principal eigenvector T
• The rank of peer i is Ti

14
What about EigenTrust?
A
9GB
9GB
1GB
10GB
B

• EigenTrust clone of PageRank
• Basic idea
• Consider recommenders reputation
• Trust matrix M
• mi,j trust of peer i to peer j (e.g download
  quantity)
• normalize each row of M
• EigenTrust vector
• The left principal eigenvector T
• The rank of peer i is Ti

1GB
C
10GB
30GB
15
False negative of EigenTrust
How the leg-hugger has high score
leg-hugger Larry

• Does the 734KB upload to Ted really matter?
• No, Ted is an irrational user
• It downloads only 124MB, but uploads 3.8TB.

16
False positive of EigenTrust(local distributor
Wayne)

• Wayne is in a satellite cluster
• Wayne uploads 290GB.
• Its EigenRank equals to a peer in majority
  community with 10GB upload
• Is it fair?
• At least, Wayne should have high rank inside the
  satellite cluster.
• We need personalized rank for each peer, e.g.
  Tit-for-Tat

5600GB
Local distributor Wayne
17
Talk organization

• The Maze p2p file sharing system
• Existing collusion behaviors
• Why simple algorithms do not work
• EigenTrust and Tit-for-Tat
• Multi-trust algorithm
• Evaluation
• Summary and Related work
• Conclusion

18
Private history Tit-for-Tat

• Idea trust peers (friends) who has helped me
  before
• Used in eMule and BitTorrent (the 2 popular P2P
  filesharing system)

19
Private history Tit-for-Tat

• Idea trust peers (friends) who has helped me
  before
• Used in eMule and BitTorrent (the 2 popular P2P
  filesharing system)
• Problem extremely small coverage

????
Limited coverage even with longer history
20
Talk organization

• The Maze p2p file sharing system
• Existing collusion behaviors
• Why simple algorithms do not work
• EigenTrust and Tit-for-Tat
• Multi-trust algorithm
• Evaluation
• Summary and Related work
• Conclusion

21
Multi-trust incentive algorithm

• Idea we need more than one tier of trust!

• get friends 1-hop friends
• build friends friend list, i.e., 2-hop friend
  list
• get friends 2-hop friends
• build 3-hop friend list

22
Multi-trust incentive algorithm

• Idea we need more than one tier of trust

• get friends 1-hop friends
• build friends friend list, i.e., 2-hop friend
  list
• get friends 2-hop friends
• build 3-hop friend list

23
Multi-trust incentive algorithm

• Idea we need more than one tier of trust

• get friends 1-hop friends
• build friends friend list, i.e., 2-hop friend
  list
• get friends 2-hop friends
• build 3-hop friend list

24
Multi-trust incentive algorithm

• Idea we need more than one tier of trust

• get friends 1-hop friends
• build friends friend list, i.e., 2-hop friend
  list
• get friends 2-hop friends
• build 3-hop friend list

25
Multi-trust incentive algorithm

• Idea needs more than one tier of trust

• get friends 1-hop friends
• build friends friend list, i.e., 2-hop friend
  list
• get friends 2-hop friends
• build 3-hop friend list

A
B
C
D
1-hop friends
E
2-hop friends
3-hop friends
F
other peers

A
C
D
E
F
B
26
Multi-trust incentive algorithm

• Idea needs more than one tier of trust

Mathematically answer use full spectrum M, M2,
M8

• get friends 1-hop friends
• build friends friend list, i.e., 2-hop friend
  list
• get friends 2-hop friends
• build 3-hop friend list

M
M2
M3
27
Multi-trust incentive algorithm
multi-trust the full spectrum incentive algorithm
Tit-for-Tat
M8?T EigenTrust
M, M2, M8
Coverage
Personalization

• Evaluation
• Coverage real trace driven simulation of one
  month
• Effectiveness statically evaluate the next 2
  weeks traffic
• Metric colluders queue position at the data
  source peer

28
Talk organization

• The Maze p2p file sharing system
• Existing collusion behaviors
• Why simple algorithms do not work
• EigenTrust and Tit-for-Tat
• Multi-trust algorithm
• Evaluation
• Summary and Related work
• Conclusion

29
Multi-trust incentive algorithmCoverage
experiment

• The coverage of M, M2 is already good enough
• We can choose using M, M2, M8

30
Multi-trust incentive algorithm Effectiveness
expr. methodology

• Setup
• Generating rank based on one months history
• Evaluate the next two weeks
• Metric
• We dont have a global rank
• Queue Position at each source peer
• Source peers who holds interested resource to me

31
Multi-trust incentive algorithm dealing with
colluders
Desirable as good as EigenTrust

• Spam account colluder
• 5/7 punish Ingrid equally
• Peer 7 punishes more in multi-trust
• Peer 4 punishes less in multi-trust since it
  downs from Ingrid

• Pair-wise colluder
• 7/9 punish Cindy equally
• 2/9 punish more in multi-trust
• Friends get ahead!

32
Multi-trust incentive algorithmsolve problems
in EigenTrust
False-negative
False-positive

• Leg-hugger
• 78 peers rank Larry lower
• 22 are still affect by super peers Ted.

• Local distributor
• Inside
• 2/3 peers promote Waynes rank
• 1/3 is too young to know Waynes good
• Outside another friend

33
Talk organization

• The Maze p2p file sharing system
• Existing collusion behaviors
• Why simple algorithms do not work
• EigenTrust and Tit-for-Tat
• Multi-trust algorithm
• Evaluation
• Summary and Related work
• Conclusion

34
Summary and Related Work
35
Conclusion

• EigenTrust and Tit-for-tat each have their own
  pitfall
• Multi-trust as a hybrid achieves better balance

36
Thank you

• QA