SybilGuard: Defending Against Sybil Attacks via Social Networks - - PowerPoint PPT Presentation

◭◭ ◮◮ ◭ ◮ Page 1 of 16 Go Back Full Screen Close Quit

SybilGuard: Defending Against Sybil Attacks via Social Networks

Haifeng Yu, Michael Kaminsky ,Phillip B. Gibbons, and Abraham Flaxman. In procedings of the 2006 conference on Applications, technolo- gies, architectures, and protocols for computer communications (SIGCOMM06), pp. 267-278, September 11–15, 2006, Pisa, Italy. Presented by Dina Adel Said

dsaid@cs.vt.edu

◭◭ ◮◮ ◭ ◮ Page 2 of 16 Go Back Full Screen Close Quit

Problem Definition

• Sybil attacks:

multiple fake identities that pretend to be multiple, distance nodes in the system.

• Problem: may out vote the honest users

in the collaborative tasks.

◭◭ ◮◮ ◭ ◮ Page 3 of 16 Go Back Full Screen Close Quit

Solutions

• Central authority

– Difficult to find a single entity that is trusted worldwide – Single point of failure – single target of denial-of-service attacks – Bottleneck of performance – The requirement of sensitive information or payment

• Binding an identify to an IP-address

– IP harvesting – co-opt a large number of end-user machines

◭◭ ◮◮ ◭ ◮ Page 4 of 16 Go Back Full Screen Close Quit

Solutions (cont.)

• Resource challenges approaches

– Posted/validated simultaneously – Adversary can have more resources than the typical user

• Network coordinates

– Can be fabricated

• Historical behavior

– Not sufficient

◭◭ ◮◮ ◭ ◮ Page 5 of 16 Go Back Full Screen Close Quit

SybllGuard

• Limits the corruptive influence of Sybil

attacks

• Build trust peer-to-peer relation between

honest nodes.

• Limit the number of attack edges be-

tween honest and malicious nodes.

◭◭ ◮◮ ◭ ◮ Page 6 of 16 Go Back Full Screen Close Quit

Social Network

• How?

– Undirected edge between two nodes if they have strong relationship – Routes are uniformly randomly generated – Convergence Property – Back-traceable property – A common edge must exist in the route be- tween verifier and suspect nodes

• Bounding Number of Attack Edges

– Malicious users can establish social trust with honest users. – If Malory convinced Alice to trust more mali- cious nodes, the number of attack edges will be the same.

◭◭ ◮◮ ◭ ◮ Page 7 of 16 Go Back Full Screen Close Quit

Bounding the number of Sybil Groups to g and Group size to w

• Why?

– Maintaining replicas of a file− > use (gw + 1) replicas – Authentication− > use (2gw + 1) nodes

◭◭ ◮◮ ◭ ◮ Page 8 of 16 Go Back Full Screen Close Quit

Bounding the number of Sybil Groups to g and Group size to w

• How?

– Limit Number of attack edges to g – Limit the No. of distinct random routes to w – Accept only one node at a given interestion point and adjacent edge

◭◭ ◮◮ ◭ ◮ Page 9 of 16 Go Back Full Screen Close Quit

Problematic Random Routes

• Causes:

– Loops – Enters a Sybil region

• Loop can only form at the starting node p = 1/d2)
• Redundancy
• Use a threshold t for number of interesting routes
• Tradeoff of t: (d = t/2)

◭◭ ◮◮ ◭ ◮ Page 10 of 16 Go Back Full Screen Close Quit

Registration of Random Path

• Tokens

– Public/Private Key authentication – Does not prevent Sybil nodes

• Shared key (symmetric ) authentication for edge

keys.

◭◭ ◮◮ ◭ ◮ Page 11 of 16 Go Back Full Screen Close Quit

Results

• Using a million-node Graph and varying

the number of attack edges

• Probability of routes remaining entirely

within the honest region is nearly 100% for majority routes

◭◭ ◮◮ ◭ ◮ Page 12 of 16 Go Back Full Screen Close Quit

Results (cont.)

• Probability of an honest node accepting

another honest node (TN) is nearly 100% for redundancy >= 10

◭◭ ◮◮ ◭ ◮ Page 13 of 16 Go Back Full Screen Close Quit

Advantages

• Decentralized Protocol
• Succeed in limiting the number of mali-

cious groups and their size

◭◭ ◮◮ ◭ ◮ Page 14 of 16 Go Back Full Screen Close Quit

Problems

• It does not limit the Sybil nodes. Increasing the

number of Sybil nodes may not significantly affect the performance while limiting their influence but it will increase computational complexity.

• The authors did not provide an experimental eval-

uation study in terms of computational complex- ity.

• The authors did not examine the probability of

accepting malicious node (FN)

• The authors limits the number of attack edges to

g by limiting the degree of the node with some constant (30).

◭◭ ◮◮ ◭ ◮ Page 15 of 16 Go Back Full Screen Close Quit

◭◭ ◮◮ ◭ ◮ Page 16 of 16 Go Back Full Screen Close Quit

Thank you dsaid@cs.vt.edu