CSCE 715Ankur Jain11/16/2010
Providing Witness Anonymity in Peer-to-Peer Systems
Bo Zhu, Sanjeev Setia and Sushil Jajodia
IntroductionDesign GoalsFrameworkSDT ProtocolAchievements of GoalsOverhead of SDTConclusion
Outline
Peer-to-Peer systemsDistributed application architecturePartitions task between peers
equivalently.E.g. – Skype, Cloud Computing, P2PTV
and many more.Fundamental Challenge
Trust relationship between peers.Several research studies.To build trust and reputation between
peers.
Introduction
ReliabilityComputing true trust value.Presence of malicious user.
AnonymityNon Identification of peers
AccountabilityIdentification of malicious peers.
Previous research focused on reliability.
Requirements for Trust Management
Overall GoalExtend P2P trust management systemsTo provide Witness Anonymity
To provide anonymity to person reporting malicious behavior.
To preserve privacy of peers.
To hide trust topology from malicious parties.
Motivation
Identity Anonymity
Backward Anonymity
Traceability
Non-slanderability.
Additional GoalsEfficiencyDecentralization.
Design Goals
System Model No Trusted Third Party.2 types of user
Offline Group Manager (OGM)User
# of adversaries less than threshold t.
Adversary Model 2 types of adversaries
Malicious userSelfish user
Will collude together to maximize the attack.
Framework
Network Model Mixnet based anonymous system
Consist of series of servers called MIXes.Associated with public keys.Receives encrypted messages.Decrypts, batches, permutes, forwards
messages.Strips off sender’s name and identifying
information.
Mechanism for monitoring claims sentIrrespective of claims being generated or
forwarded.
Framework
SDT – Secure Deep ThroatProvide anonymity and accountability together.Include tracing mechanism to identify user.
4 step procedureSetupRegistrationClaim BroadcastingPublic Tracing
Modes of OperationActive: Real Time requirements.Passive: Not strict Real Time requirements.
SDT Protocol
OGM generates public and secret keys.
Identification list (LIST) initially empty.
Define tag basesUsed in claim broadcasting
To create anonymous claims.
Only one per type of misbehavior per user.
SDT Protocol – Setup
User contacts OGM.
User selects identity.
Check its availability.
User obtains a member public/secret key pair.
OGM adds a new entry to LIST.
OGM select s items from LIST and sends it to user.
User sends confirmation for key pair and LIST items received.
SDT Protocol – Registration
User maintain two databases. Maintains claim sent by herself.Maintains claim received from other user.
On detecting malicious behaviorChecks database for previous entries for
same type of behavior.If not found generates new claim using tag
base.Broadcast through anonymous
communication system.Also stores claim in database.
SDT Protocol – Claim Broadcasting
On receiving claim Checks whether entry for that claim is
present or not.If yes, then drops the claim.If not check its validity and stores the claim.Also forwards it in the system.
Initializing Public TracingUser finds t claims.Checks distinctness of all t claims.Generates a message including t claims and
broadcast it to network
SDT Protocol – Claim Broadcasting
Check for entries in databases.
If found broadcast two entries as proof to disclose the identity of malicious user.
If no entries found broadcast message NO-ONE.
After receiving NO-ONE message other repeat the steps in their local LIST.
SDT Protocol – Public Tracing
Used when real time requirement is not critical.
Achieve better efficiency.
Changes in claims broadcastingClaims regarding malicious behavior not sent
immediately.Sent these claims only when queried about the
behavior of user.
Public tracing will performed on all claims to prevent multiple claims from an adversary.
SDT Protocol – Passive Mode
Peer forwards claim with a probability.Instead of flooding entire network.Lower the probability, lower is the number
of peers storing the claim.Lower is the probability that one peer
stores every t distinct claims.Require more number of witnesses in this
case.Also non zero probability that adversary
may escape disclosure.
SDT Protocol – Probabilistic Forwarding
Identity AnonymityMay be broken using Traffic Analysis or Protocol
Analysis
Traffic Analysis is prevented by Mixnet based communication system.
Protocol Analysis is also hard to perform
No public key in claim broadcasted
All parameter are calculated using discrete algorithm so very robust against brute force attack.
Achievement of Goals
Backward AnonymityAdversaries can compromise multiple peers.
Claim does not provide information regarding identity.
No way to differentiate the user on basis of claims.
Also ensured when OGM and adversaries are in contact
User’s secret key is only known to user.
No way to extract secret key from OGM.
Achievement of Goals
TraceabilityGood peers need to find a valid record of
adversary from LIST.
LIST items are distributed among different peers.
Probability of all copies controlled by adversary group is very small.
Achievement of Goals
Non-SlanderabilityMax number of claims sent by adversaries
against a user
Total number of adversaries which is less than t.
Adversaries cannot collect enough claims to remove good user from the system.
Achievement of Goals
Distributed storage of LISTOGM maintains LIST offline.LIST is stored in distributed form.Peers do not have knowledge of LIST items
with other peers.Helps in detecting a adversary even if
adversary is controlling the majority of LIST.
Overhead of SDT
Communication CostsMajor cost is forwarding claims.
Implemented using elliptic curve or hyper elliptic curve over a finite field.
Claim size not more than 409 bytes.
LIST distribution another cost.
Smaller the LIST, higher probability of message broadcast while tracing.
Overhead of SDT
Storage RequirementsFor cryptographic keys, LIST and local
databases.Storing personal keys and public key of OGM.
Only small part of the entire LIST.
Very small database requirement in passive mode.
A probabilistic forwarding approach may reduce database space in active mode.
Overhead of SDT
SDT provide witness anonymity to users reporting malicious behavior.
Two modes of operation: Active and Passive.
Overhead is acceptable in peer-to-peer systems.
Conclusion
Questions