CT-ISG Scalable Byzantine Replication under Attack

Johns Hopkins University, Baltimore MD

Investigators

Abstract

As network environments become increasingly hostile, even well protected distributed information systems, constructed with security in mind, are likely to be compromised. Hence, architecting large-scale distributed systems that function correctly and provide adequate performance even when parts of them are compromised is one of the most important challenges. Byzantine replication is emerging as a promising direction to mitigate server compromise. Experience with Byzantine replication protocols reveals considerable shortcomings in the underlying theoretical foundation. Namely, existing fault models, metrics, and correctness criteria used to reason about and construct Byzantine replication algorithms, fail to capture properties that manifest themselves in wide-area environments. Since all existing Byzantine replication algorithms were designed to meet the standard safety and liveness criteria, they all exhibit critical vulnerabilities not covered by the standard models. This project will develop the theoretical foundation, architectural framework, and algorithmic techniques for a scalable wide-area Byzantine replication system that provides strong performance guarantees under attack. This includes: (1) expanding the existing theoretical fault models to better encapsulate the unique characteristics and performance vulnerabilities associated with scalable wide-area Byzantine replication systems; (2) defining useful metrics for evaluating and comparing different architectures, configurations, and algorithms with a focus on their performance under sophisticated attacks; (3) developing an architectural framework for scalable Byzantine replication that can be customized to topology, performance, and resiliency requirements of specific wide-area systems; and (4) developing specific algorithms for wide-area environments that provide strong performance guarantees under attack.