On generalized low-rate denial-of-quality attack against Internet services

Abstract

Low-rate Denial of Quality (DoQ) attacks, by sending intermittent bursts of requests, can severely degrade the quality of Internet services and evade detection. In this paper, we generalize the previous results by considering arbitrary attack intervals. We obtain two sets of new results for a web server with feedback-based admission control. First, we model the web server under the attack as a switched system. By proving the Lyapunov and Lagrange stability of the system, we show that the admission rate can always be throttled to a bounded low value. Second, we investigate the worst impacts of a DoQ attack by optimizing a utility function for the attacks. As a result, we obtain for the first time optimal attack patterns for both periodic and aperiodic attacks. Extensive simulation results also agree with the analytical results.