Self-protecting Systems -analysis and Comparison (by Nikolaos Michalakis)

This document analyses five systems that protect themselves from failures and malicious attacks by coupling and automating protection with their regular operation. We call such systems self-protecting. The goal of designing self-protecting systems is to reduce the trust placed on humans or external protection mechanisms and the corresponding manageability costs. The systems analyzed are: Practical Byzantine Fault Tolerance, LOCKSS, EigenTrust, SSM and Rx. Our findings are shown in the table below. From our analysis we extract two important design techniques for self-protection: active selfrepair and non-deterministic defense. The systems are later compared with respect to these two techniques.