Date of Award

Fall 2011

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Electrical Engineering - (Ph.D.)

Department

Electrical and Computer Engineering

First Advisor

Roberto Rojas-Cessa

Second Advisor

Nirwan Ansari

Third Advisor

Sotirios Ziavras

Fourth Advisor

Yanchao Zhang

Fifth Advisor

Guiling Wang

Abstract

Peer-to-peer (P2P) networking enables users with similar interests to exchange, or obtain files. This network model has been proven popular to exchange music, pictures, or software applications. These files are saved, and most likely executed, at the downloading host. At the expense of this mechanism, worms, viruses, and malware find an open front door to the downloading host and gives them a convenient environment for successful proliferation throughout the network. Although virus detection software is currently available, this countermeasure works in a reactive fashion, and in most times, in an isolated manner. A trust management scheme is considered to contain the proliferation of viruses in P2P networks. Specifically, a cooperative and distributed trust management scheme based on a two-layer approach to bound the proliferation of viruses is proposed. The new scheme is called double-layer dynamic trust (DDT) management scheme. The results show that the proposed scheme bounds the proliferation of malware. With the proposed scheme, the number of infected hosts and the proliferation rate are limited to small values. In addition, it is shown that network activity is not discouraged by using the proposed scheme. Moreover, to improve the efficiency on the calculation of trust values of ratio based normalization models, a model is proposed for trust value calculation using a three-dimensional normalization to represent peer activity with more accuracy than that of a conventional ratio based normalization.

Distributed network security is also considered, especially in P2P network security. For many P2P systems, including ad hoc networks and online markets, reputation systems have been considered as a solution for mitigating the affects of malicious peers. However, a sybil attack, wherein forging identities is performed to unfairly and arbitrarily influence the reputation of peers in a network or community. To defend against sybil attack, each reported transaction, which is used to calculate trust values, is verified.

In this thesis, it is shown that peer reputation alone cannot bound network subversion of a sybil attack. Therefore, a new trust management framework, called Sybildefense, is introduced. This framework combines a trust management scheme with a cryptography mechanism to verify different transaction claims issue by peers, including those bogus claims of sybil peers. To improve the efficiency on the identification of honest peers from sybil peers, a k-means clustering mechanism is adopted. Moreover, to include a list of peer’s trustees in a warning messages is proposed to generate a local table for a peer that it is used to identify possible clusters of sybil peers. The defensive performance of these algorithms are compared under sybil attacks. The performance results show that the proposed framework (Sybildefense) can thwart sybil attacks efficiently.

Share

COinS