- 期刊
A Statistical P2P Botnet Detection Resilient to Mimicry Attacks
摘要
Botnet technology has continued to evolve rapidly, making detection a very challenging problem. P2P botnets are more dangerous and resistant than all emerged botnets due to their distributed architecture. The most proposed P2P botnet detection schemes are designed, relying on the statistical behavior of bots. However, considering the adversarial nature of the botnet detection problem, the bots can be designed to mimic normal behavior and fly under the radar. Thereupon, designing a P2P botnet detection system resilient to the mimicry attack is paramount. In this paper, we implement a mimicry P2P botnet to investigate the resiliency of existing P2P botnet detection schemes. Furthermore, a statistical feature set is proposed to leverage botnets' inherent properties. Our experimental results showed that the proposed feature set is resilient to mimicry attacks and can detect P2P bots with high accuracy.