Yitzchak Rosenthal

1
P2P Mechanism Design Incentives in Peer-to-Peer
SystemsPaper By Moshe Babaioff, John Chuang and
Michal Feldman

• Yitzchak Rosenthal

2
Types of P2P networks

• P2P network applications
• File downloading (e.g. BitTorrent, Gnutella,
  etc.)
• Video streaming
• MANETs

3
P2P Issues

• Private Information
• Many P2P protocols require clients to divulge
  private information.Examples
• Amount of bandwidth a client has for uploading
  files.
• List of files/data client has for uploading
• Clients may choose NOT to divulge private
  information in order to exploit the network for
  its own gain.
• Free Riding
• Peers try to get use OF network without providing
  services TO network(e.g. downloading data from
  peers without uploading to peers)
• Whitewashing
• If multiple identities can be created for free
  then an evil user can destroy an identity once
  it has been recognized as not following the rules
  and exploiting the network
• Sybil Attacks
• Multiple IDs by same user that collude with each
  other

4
Addressing the problem through Incentives

• Provide incentives to peers to follow the rules
• Types of incentives
• Currency (CUR) - Mojonation
• Peers earn currency when providing TO the
  network.
• The currency can be spent in order to get
  services/data FROM the network
• Reputation (REP) - KaZaA
• Peers get a better reputation when they provide
  TO the network
• Peers with better reputation get better download
  speed
• Barter (BAR) - BitTorrent
• Scalable - doesnt keep state information (CUR
  and REP do)
• Files are broken into many equal size chunks
• seeder peer distributes DIFFERENT chunks to
  many different peers
• Peers who have a chunk exchange with peers who
  have other chunks.

5
Case study File Sharing Networks
6
One shot game

• In a ONE SHOT GAME - free Riding is the dominant
  strategy
• Similar to one shot Prisoners Dilemma (PD) where
  dominant strategy is to defect.
• No downsides for cheating
• No loss of reputation
• No way to spend any income

7
Other approaches

• Direct reciprocity can be better, but
• In large population, effect of direct reciprocity
  is diluted since the odds of interacting again
  with same peer is low (Friedman, et al) (See next
  slide)
• Enforce direct interaction with limited number of
  peers (BitTorrent)
• Reputation systems introduces state may not
  scale as well
• How to deal with newcomers
• Dissuade whitewashing by
• Cooperate with strangers with a fixed
  probability, p, is not robust against white
  washers
• Better approach is adjust p based on frequency of
  past cooperation with strangers. This works
  better for a small turnover rate.

8
Dilution of effects of direct reciprocity with
large population.
9
Reputation
10
Reputation

• Areas that reputation work
• Evolutionary biology
• Online marketplaces (e.g. eBay)
• FileSharing - eg. KaAzA files who upload have
  better reputation scores and get higher priority
  when downloading
• Eigentrust algorithm
• Uses transitive trust relationships to
  aggregates local trust values to form global
  trust values
• Similar to page rank in Google
• Credence algorithm
• Extends trust from peers to objects in p2p
  system to defend against pollution and poisoning

11
Minimalist P2P model (no reputation)

• Each peer i has type ? generosity amount that
  peer will contribute to system
• x of contributors to system
• Contribution cost per peer 1/x
• Decision of rational peer
• See graph on next slide

12
Miminimalist P2P model - costs

• Y axis is of contributors to system
• X axis is generosity level
• x1, x2 (on Y axis) and zero are equilibria of
  system
• X2 is NOT a stable equilibrium
• Generosity ? is uniformly distributed beween 0
  and ?m.
• Straight line is CDF of percent of peers who will
  contribute at a certain price level.
• Curved line is the model of the cost per
  contributor.

13
Solve for x1 and x2

• Solve for

14
Benefits

• Benefit proportional to contribution level a
• Performance of System
• Ws ax (1/x)x ax -1 (note 1/x is used
  instead of 1/ ? )
• System will still collapse if maximal generosity
  is low

15
Reputation system

• Catch free riders with probabilty, p, an
  eliminate free riders from systemORcatch free
  riders with probability 1 and peanalize free
  riders with (1-p) times reduced service of
  contributor
• Load placed on system decreases to
• So contribution cost becomes

16
Analysis with reputation

• Q individual benefit
• R reduced contribution cost
• T threat
• Contributor performance Q R
• Free Rider performance
• System Performance

17
Analysis of Barter Based System (BitTorrent)
18
Principal Agent Model

• N set of agents
• n of agents
• Ai 0,1 set of possible actions for each
  agent, i ? N
• a specific agent
• Set of n agents, N