Queue Management as a DoS counter-measure?

Abstract. In this paper, we study the performance of timeout-based queue management practices in the context of flood denial-of-service (DoS) attacks on connection-oriented protocols, where server resources are depleted by uncompleted illegitimate requests generated by the attacker. This includes both crippling DoS attacks where services become unavailable and Quality of Service (QoS) degradation attacks. While these queue management strategies were not initially designed for DoS attack protection purposes, they do have the desirable side-effect or providing some protection against them, since illegitimate requests time out more often than legitimate ones. While this fact is intuitive and wellknown, very few quantitative results have been published on the potential impact on DoS-attack resilience of various queue management strategies and the associated configuration parameters. We report on the relative performance of various queue strategies under a varying range of attack rates and parameter configurations. We hope that such results will provide usable configuration guidelines for end-server or network appliance queue hardening. The use of such optimisation techniques is complementary to the upstream deployment of other types of DoS-protection countermeasures, and will probably prove most useful in scenarios where some residual attack traffic still bypasses them