PFS: Probabilistic filter scheduling against distributed denial-of-service attacks

Abstract

Distributed denial-of-service (DDoS) attacks continue to pose an important challenge to current networks. DDoS attacks can cause victim resource consumption and link congestion. A filter-based DDoS defense is considered as an effective approach, since it can defend against both attacks: victim resource consumption and link congestion. However, existing filter-based approaches do not address necessary properties for viable DDoS solutions: how to practically identify attack paths, how to propagate filters to the best locations (filter routers), and how to manage many filters to maximize the defense effectiveness. We propose a novel mechanism, termed PFS (Probabilistic Filter Scheduling), to efficiently defeat DDoS attacks and to satisfy the necessary properties. In PFS, filter routers identify attack paths using probabilistic packet marking, and maintain filters using a scheduling policy to maximize the defense effectiveness. Our experiments show that PFS achieves 44% higher effectiveness than other filter-based approaches. Furthermore, we vary PFS parameters in terms of the marking probability and deployment ratio, and find that 30% marking probability and 30% deployment rate maximize the attack blocking rate of PFS.