Attack resistant cache replacement for survivable services

Many distributed services are susceptible to attacks by malicious clients that can significantly degrade their performance. Scalable distributed services make use of a variety of techniques which are vulnerable to such attacks. We explore the survivability of services when attacks target the scalability techniques employed by the services. In particular, we explore how the effectiveness of caching can be degraded when malicious clients manipulate cache management algorithms. We present an attack resistant replacement algorithm and show that it is much more effective in dealing with attacks compared to other widely deployed replacement algorithms.