Using Attestation to Lift Crash Resilience to Byzantine Resilience

This paper explores the use of attestation protocols as Byzantine failure detectors. An attestation protocol enables one node in a distributed system to obtain enough information about other nodes to detect malicious compromises. By filtering network communication, channels to Byzantine nodes are made to appear crashed. Distributed algorithms that tolerate channel failures are thus transformed into ones that tolerate Byzantine failures. Erlang modules to support filtering and attestation have been written, including a partial Trusted Platform Module (TPM) interface. A demonstration prototype for a leader election algorithm is in progress.