P2P Security - Reputation and Credibility

P2P Security - Reputation and Credibility

1.0

07/October/2003

Niels Olof Bouvin

Aarhus Universitet

Recapture from last time:

·  Cryptography enables (to a high degree) secure communication

·  With asymmetric cryptography (public/private keys) messages can be free from eavesdropping and tampering

·  Thus, we have the basics for building a secure infrastructure

·  This however still leaves the question: Who can you trust?

Reputation and Credibility

·  Introduction

·  Reputation and moderation systems

Introduction

Who can you trust?

·  Trust is inherent in most of our social interactions

·  Generally we have a pretty good idea whom to trust and whom not to trust

·  People (and companies) earn and loose our trust over time

·  If Alice trusts Bob, and Bob trusts Daniel, Alice is more likely to trust Daniel

o  and if Bob does not trust Carol, Alice is unlikely to trust Carol

·  Trust is hard to earn - easy to loose

o  a powerful incentive to be nice

Trust in a computer setting

·  In real life, we have a number of mechanisms for handling trust, and we are usually dealing with people we know

·  On the Internet, these mechanisms break down

o  "on the Internet, no one knows you're a dog"

·  You can establish trust across the Internet

o  and the techniques outlined last time can help you ensure that you are dealing with the person you expect

Some types of attack on the Internet

·  Pseudospoofing

o  using a number of aliases to influence a situation (e.g. voting)

o  on-line identities are not tied to real world identities - people cannot be prevented from using various aliases

·  Denial of Service

o  request a service so much it becomes unavailable

o  can be handled (to a degree) through caching and micropayment

·  Flooding

o  filling a system with garbage drowning out the valid content

o  various defences: micropayment or dropping unpopular content

Reputation and moderation systems

Usenet

·  People on Usenet earn their reputations by their postings

·  Some are respected as authorities - others are reviled as kooks, flamers, spammers, and trolls

·  Especially trolls will often change their identities as they invariantly are kill filed by their unwilling audience

o  uncovering these new identities is a pastime of some newsgroup residents - but they are of course always one step behind the troll

Usenet - anti-{kook,flamer,troll} weaponry

·  Most news readers have 'kill file' functionality

o  the user adds an unwanted poster to his/her kill file and the unwanted one is then automatically filtered by the news reader

o  disadvantage: binary, requires user intervention. Per user.

·  Some news readers (Gnus) use adaptive scoring

o  postings and users are awarded scores depending on the reader's actions

o  advantage: much more flexible

o  still not shared - does not work for newsgroups only visited seldom or for the first time

·  Apart from adaptive scoring, scoring can be explicitly designated, so that specific subjects or authors are scored up or down. Scoring can depend on arbitrary computations on the postings

o  disadvantage: complex

Usenet - moderation

·  Some newsgroups become too cluttered by spam or flame wars

·  Either the newsgroup is left to deteriorate (with the regulars moving elsewhere) or the newsgroup becomes moderated

·  A number of trusted regulars are elected as moderators by the newsgroup

·  A posting must hence forward be approved by a moderator, before it appears on the newsgroup

·  Problems:

o  moderation is hard work

o  moderators can misuse their power

o  moderation of newsgroup can be bypassed

Usenet - NoCeM and GroupLens

·  Moderators do a great job, but

o  not all newsgroups are moderated

o  you may not agree with all the moderators' decisions

·  NoCeM

o  opt-in moderation - if you trust a moderator, you can subscribe to his/her/its decisions

·  GroupLens

o  moderation done by people you agree with

o  you assign scores to individual postings, and thus build a profile

o  the moderation done by people with similar profiles are then used to score the postings they see

Slashdot

·  Purpose

o  moderate postings on the discussion boards by scoring interesting posts up and scoring uninteresting/inflammatory posts down

·  Moderation were traditionally handled by a group of trusted moderators

o  problem: way too many posts for the moderators to handle

o  misuse of moderator power (who watches the watchers?)

Slashdot - meta-moderation

·  All (registered) users in good standing occasionally gets (limited) moderation power

·  Moderators cannot themselves participate in the discussions they moderate

·  The moderations are meta-moderated by all registered users

·  Bad moderators will not moderate again (or very rarely)

Amazon

·  Purpose:

o  the recommendation of books, you're likely to like (and buy)

·  Based on the buying habits of the Amazon customers

·  You're (presumed) likely to buy books bought by people, who buy the books you buy

·  This works surprisingly well

eBay

·  How can you trust a stranger with your money?

·  Sellers and buyers mutually rate each other after each transaction

·  So, you don't buy stuff from sellers with bad reputations

·  Problem:

o  you can build up a good reputation over many smaller transactions and then cheat people on high price deals

Advogato

·  System to establish the reputation of open source programmers

o  users assign trust to each other - their own reputation determines how much their trust is worth to others

o  pseudospoofing attacks become difficult as new users have zero trust to give to others

Micropayment - making cheating unprofitable

·  DoS attacks becomes infeasible if attackers must pay for the used resources

·  Cash payment

o  using a number of sophisticated cryptographic methods

o  very much an ongoing research topic

·  Proof of Work (POW)

o  performing some non-trivial work to gain access to a resource

o  typically variation of cryptographic hash collision calculation

Summary

·  Existing reputation systems provides users a modicum of quality of service

·  However, they are not tamper proof and may be misused

·  Even if the moderators have integrity, you may not agree with them

·  Systems must have checks and balances in order to prevent abuse

P2P systems

·  Reputation and security in P2P systems

·  Damiani et al.

·  Gupta et al

Reputation and security in P2P systems

Security?

·  File sharing networks such as Gnutella do not have much protection against malicious users or dangerous content

·  Some systems are more or less censorship proof (such as Freenet), but what about bad content and users?

·  Or flooding and DoS?

·  A central authority is not available in the fully distributed systems

·  Most users are anonymous

·  Who can you trust?

Damiani et al.

Distributed reputation in Gnutella using XRep

·  Basic idea: collect other peers' votes on the reputation of resources and other peers

·  Dangers:

o  Malicious users may distribute dangerous content

o  Other users may (unwittingly) distribute dangerous content

o  Malicious users may lie about themselves or others

Gnutella revisited - how does it work again?

·  A Gnutella peer starts out with a number of peers from "somewhere"

·  It can query these peers for the peers they know, and thus build a list of potential peers to contact

·  Queries are broadcast to all known peers who in turn call all their peers and so on

·  Queries has a unique ID (128 bit) and a TTL (time to live). This ensures that peers do not retransmit the same query twice and that queries eventually die out

·  Peers remembers received and transmitted queries and whence they came

·  If a query match is found, the response (containing the query and the host address) is returned following the query route back to the originator

·  The originator will (presumably) received a number of hits and then contact a host directly for downloading

Extensions to the Gnutella system

·  Peers are known as "servents" (server/clients)

·  All servents have (public,private) keys

·  All servents have a unique ID based on a hash of their public key

·  All resources have a digest - a cryptographic hash

o  similar to Freenet...

·  All servents have resource repositories, consisting of

o  (resource_id, value)

o  value: + or -

·  All servents have servent repositories, consisting of

o  (servent_id, num_plus, num_minus)

Resource Searching in XRep

·  Queries are broadcast in normal fashion

·  Query replies additionally contain

o  servent_id

o  digests for the matching resources

Resource searching

Resource selection and vote polling

·  The user decides on a resource

·  The servent queries peers about the resource and servents offering it.

·  The vote polling query consists of

o  the servent's generated poll public key

o  the resource digest

o  the servent_id for the supplying servent

·  Other servents check their repositories and return their knowledge - encrypted with the public key

Resource selection and vote polling

Vote evaluation

·  Votes are decrypted

·  To prevent cheating by pseudospoofing IPs are collapsed and random voters queried to confirm their vote

Vote evaluation

Best servent check

·  Based on the vote, the best servent is selected

·  To prevent overloading the best servent, it is contacted for the resource digest (thus confirming the existence and identity of the resource)

Best servent check

Resource downloading

·  A servent is decided upon, and downloading commences

·  After download, the resource is checked against the digest and its repositories are updated

Resource downloading

Summary

·  XRep makes it more difficult to pseudospoof by requiring different IPs (this can of course be circumvented)

·  Again we see the usefulness of the cryptographic hash

·  Dangerous content gets a bad score in this system

o  making sure we don't download it - even from good servents

·  Adds even more traffic to Gnutella...

Gupta et al

Reputation relying on a network of reputation agents

·  Basic concept:

o  peers calculates scores after each transaction

o  these scores are collected by Reputation Computation Agents (RCA)

·  Reputation scores depend on

o  Behaviour

§  Content search contribution

§  Content download contribution

o  Capability

§  Processing power

§  Bandwidth

§  Storage capacity

§  Memory

·  Participation is voluntary

·  All peers, including RCA, have (public,private) keys

Overall approach

Debit-Credit Reputation Computation

·  Credit based on

o  how often the peer responds to queries (QRC)

o  how much the peer uploads compared to its bandwidth (DC)

o  how much the peer shares (SC)

·  Debit based on

o  how much the peer downloads (DD)

Credit Only Reputation Computation

·  No debits, but scores deteriorates over time

RCA in action

·  Peers will regularly contact the RCA to update their score

·  A peer sends (encrypted) Proofs of Processing consisting of

o  (requester_id, query_keywords, query_size, time_stamp, self_id)

·  The RCA returns a new (encrypted) score to the peer

·  The RCA maintains a history of PPs to ensure that work is not credited twice

·  Debits are stored until a credit PP comes in (to prevent peers from throwing debits away)

Downloading - DCRC

·  The requester and sender exchange receipts for the download:

o  the receipt is signed before transfer

Downloading - CORC

·  Receipt sent after download

Summary

·  Avoids overloading (already busy) Gnutella protocol

·  At the cost of introducing a whole new layer (assumed not malicious!)

·  Defence against a group of malicious peers?

o  malicious peers could just claim to exchange large files

o  this can be handled through Advogato like techniques

Conclusions

·  Reputation and trust on the Internet is not easy

·  A number of good techniques exist - often based on a central authority

o  but can you trust the moderators?

·  P2P makes everything worse

o  no central authority

·  Trust becomes an issue of trusting well-behaving peers and their judgement

o  pretty much like real life

Created by JackSVG