Date Awarded

2010

Document Type

Dissertation

Degree Name

Doctor of Philosophy (Ph.D.)

Department

Computer Science

Advisor

Haining Wang

Abstract

Message systems, which transfer information from sender to recipient via communication networks, are indispensable to our modern society. The enormous user base of message systems and their critical role in information delivery make it the top priority to secure message systems. This dissertation focuses on securing the two most representative and dominant messages systems---e-mail and instant messaging (IM)---from two complementary aspects: defending against unwanted messages and ensuring reliable delivery of wanted messages.;To curtail unwanted messages and protect e-mail and instant messaging users, this dissertation proposes two mechanisms DBSpam and HoneyIM, which can effectively thwart e-mail spam laundering and foil malicious instant message spreading, respectively. DBSpam exploits the distinct characteristics of connection correlation and packet symmetry embedded in the behavior of spam laundering and utilizes a simple statistical method, Sequential Probability Ratio Test, to detect and break spam laundering activities inside a customer network in a timely manner. The experimental results demonstrate that DBSpam is effective in quickly and accurately capturing and suppressing e-mail spam laundering activities and is capable of coping with high speed network traffic. HoneyIM leverages the inherent characteristic of spreading of IM malware and applies the honey-pot technology to the detection of malicious instant messages. More specifically, HoneyIM uses decoy accounts in normal users' contact lists as honey-pots to capture malicious messages sent by IM malware and suppresses the spread of malicious instant messages by performing network-wide blocking. The efficacy of HoneyIM has been validated through both simulations and real experiments.;To improve e-mail reliability, that is, prevent losses of wanted e-mail, this dissertation proposes a collaboration-based autonomous e-mail reputation system called CARE. CARE introduces inter-domain collaboration without central authority or third party and enables each e-mail service provider to independently build its reputation database, including frequently contacted and unacquainted sending domains, based on the local e-mail history and the information exchanged with other collaborating domains. The effectiveness of CARE on improving e-mail reliability has been validated through a number of experiments, including a comparison of two large e-mail log traces from two universities, a real experiment of DNS snooping on more than 36,000 domains, and extensive simulation experiments in a large-scale environment.

DOI

https://dx.doi.org/doi:10.21220/s2-91s6-5e87

Rights

© The Author

Download

Share Feedback

Share

COinS