Byzantine Replication for Trustworthy Systems

University Of Texas At Austin, Austin TX

Investigators

Abstract

Proposal number: 0430510 TITLE: Byzantine Replication for Trustworthy Services Principal Investigator: Lorenzo Alvisi In a world where economics dictates that few components be rigorously tested or verified, methods for building trustworthy systems from untrustworthy components are essential. An attractive approach toward managing the complexity inherent in building trustworthy distributed systems consists in modeling a compromised component as faulty according to the Byzantine failure model---the weakest of all failure models, which allows faulty components to deviate arbitrarily and maliciously from their correct specification. This research explores the feasibility of this approach, both with respect to its fundamental assumptions and to its engineering viability. On the first front, the focus is on (1) the challenge of conjugating fault-tolerance and privacy by developing a firewall with formally verifiable privacy guarantees and (2) the establishment of a solid, quantitative basis for measuring failure independence of replicas against security attacks in Byzantine fault-tolerant architectures. On the second front, the emphasis is on exploring novel ways to implement Byzantine services that provide low latency, high throughput, and can be quickly and unobtrusively reconfigured to improve their performance in response to changes in the environment in which they operate. Addressing these issues successfully will enable a strategy for assembling untrustworthy components to obtain trustworthy systems. More broadly, the proposed research will impact both graduate and undergraduate curriculum at UT Austin and will contribute towards creating a UT Center for Information Security.