|M.Sc Thesis||Department of Computer Science|
|Supervisor:||Prof. Idit Keidar|
One of the most devastating security threats faced by a distributed system is a denial of service (DoS) attack, in which an attacker makes a system unresponsive by forcing it to handle bogus requests that consume all available resources. In a distributed denial of service (DDoS) attack, the attacker utilizes several computers as the source of a DoS attack, in order to increase the attack strength. In 2003, approximately 42% of U.S. organizations, including government agencies, financial institutions, medical institutions and universities, were faced with DoS attacks. That year, DoS attacks were the second most financially damaging attacks (65 million USD), only short of theft of proprietary information (70 million USD), and far above other attacks (0.07-27 million USD). Therefore, coping with DoS attacks is essential when deploying services in a hostile environment such as the Internet.
In this thesis, we propose a framework and methodology for quantifying the effect of DoS attacks on a distributed system. We present a systematic study of the resistance of gossip-based multicast protocols to DoS attacks. We show that even distributed and randomized gossip-based protocols, which eliminate single points of failure, do not necessarily eliminate vulnerabilities to DoS attacks. We propose Drum (DoS-Resistant Unforgeable Multicast) -- a simple gossip-based multicast protocol that eliminates such vulnerabilities. We show, using asymptotic closed-form analysis, simulations, and detailed analysis, that Drum survives severe DoS attacks, while traditional gossip-based protocols collapse even under mild attacks.