Tolerating denial-of-service attacks using overlay networks: impact of topology

Proxy-network based overlays have been proposed to protect Internet Applications against Denial-of-Service attacks by hiding an application's location. We study how a proxy network's topology influences the effectiveness of location-hiding. We present two theorems which quantitatively characterize when proxy networks are robust against attacks (attackers' impact can be quickly and completely removed), and when they are vulnerable to attacks (attackers' impact cannot be completely removed). Using these theorems, we study a range of proxy network topologies, and identify those topologies favorable for location-hiding and resisting Denial-of-Service attacks. We have found that popular overlay network topologies such as Chord [25], which has been suggested for location-hiding, is in fact not a favorable topology for such purposes; we have also shown that CAN [21], a less popular overlay network, can be a good topology for location-hiding. Our theoretical results provide a set of sound design principles on proxy networks used for location-hiding.