Proposal Pollution prevention in the P2P file sharing system

1
ProposalPollution prevention in the P2P file
sharing system

• Presenter Elaine

2
Motivation

• P2P traffic has dominated 60 traffic in the
  internet, P2P file-sharing is an important
  application.
• Recently, many existing works have shown that
  network is rife with deliberate polluted files
• Definition of polluted file
• File content does not match its file description

3
Motivation

• Application environment description
• A P2P file-sharing application with search
  capability
• File-sharing apps use meta-data for searching
• Content Hash
• Response result list
• For a given file A

Version of copies H1
40(P2,P7.P80,P91,P102) H2
23(P3,P5.P33,P54..)
Hn 2(P10,P17)
4
Related work

• Different types of pollution attack
• Decoy injection Meta data is the same, H is
  different
• File content is damaged or not match
• Hash corruption H is the same, but content is
  polluted
• Two different files could be maliciously hashed
  to the same hashed ID, dangerous especially when
  parallel downloading

5
Related work

• Peer-reputation systems exist.
• Based on the peers history of uploads
• Eigen-trust
• Even downloading from trusted peer, still cant
  guarantee for a non-polluted file
• User awareness
• User slackness

6
Related work

• Object reputation system
• Credence
• The first object-reputation system
• Voting after each object downloading? Issuing a
  vote-gather query ? Evaluating the object
  reputation.
• Two database
• Vote database
• Correlation table

7
Related work

• Credence
• Hash corruption Mechanism still can not be avoid
  because it didnt verify for the source.
• Disadvantages
• Votes database could be costly
• The correlation is not accurate if two peers
  didnt download enough common objects.

8
Problem Definition

• The best way to prevent the spreading of
  pollution is to
• Select a non-polluted file first
• Then select the trust peers to download

Version of copies H1
40(P2,P7.P80,P91,P 102) H2
23(P3,P5.P33,P54..)
Hn 2(P10,P17)
9
Idea

• Designing a robust pollution-prevention system
• Mechanism operations
• Vote after downloading each object
• Calculate each peers reputation periodically
• Searching for object and collecting votes
• Calculate objects reputation before downloading
  and select peers to download from.

10
Calculate each peers reputation

• Each time peer i download a file from peer j, it
  may rate the transaction as positive or negative
  value
• sij sat(i, j) - unsat(i, j)
• Transitive trust calculated periodically

11
Searching for object and collecting votes
Query for vote Query for object
12
Select object and trusted peers to download

• Weigh collecting votes by the trust value to the
  voter
• Select a non-pollution version
• Select a group of trusted peers to download from

13
Experimental plan

• Compare with existing strategy of
• Peer reputation system
• Object reputation system (Credence)
• random, redundant best, redundant random
  downloading
• Metrics
• From user perspective
• The necessary time for downloading a clean file
• From network perspective
• The amount of traffic generated by the
  transmission of polluted files
• The pollution level varies with time, and the
  pollution level at the steady state
• Pollution level The ratio of good copies and bad
  copies in the network
• Human factor
• User awareness
• User slackness
• Willingness to vote