Fault Detection for Message Passing Systems

Many crucial dependable and secure services including atomic commitment, consensus and group membership, and middleware services (such as replica, communication and transaction services) use fault detectors. Through the use of fault detectors, the overlying service can be exempted from failure treatment and synchronization requirements. Fault detection is essential for proving that the services carried out are correct. In this paper, we first identify the necessary conditions to detect faults in a message passing system where multiple disjoint paths exist between each pair of endpoints. We then present the first fault detection protocol capable of detecting message meta-data modification in the presence of various message interferences in addition to other faults including omission faults, message replay and spurious messages using disjoint paths, where paths with faults are not known a priori. In addition, it authenticates message origins allowing Sybil attacks to be detected, identifies faulty paths, and classifies faults in the presence of multiple messages sent by various system processes. We establish the completeness and soundness properties of the proposed algorithm, i.e., it detects each considered fault and each detected fault is an actual fault, respectively. We also show that our algorithm does not yield a significant packet size and delay overheads. The algorithm shows the viability of the use of disjoint paths in fault detection.