[Back]

J. Widder, G. Gridling, B. Weiss, J. Blanquart:
"Synchronous Consensus with Mortal Byzantines";
Talk: Dagstuhl Seminar 06371. From Security to Dependability, Dagstuhl (invited); 2006-09-10 - 2006-09-15.

It is often argued that the Byzantine failure model captures all kinds
of reasons for components to fail, even intentional faults like
intrusions. An important question is whether this is the only possible
model, or whether even more restricted failure models could be
suitable in order to address both reliability and safety
requirements. In order to give some hint of what can be done from the
reliability side, I present some recent work on novel fault models:

I will consider the problem of reaching consensus in synchronous
systems under a fault model whose severity lies between Byzantine and
crash faults. For these ``mortal" Byzantine faults, one assumes that
faulty processes take a finite number of arbitrary steps before they
eventually crash.

I present a consensus algorithm that tolerates a minority of faulty
processes; i.e., more faults can be tolerated compared to classic
Byzantine faults. We also show that the algorithm is optimal regarding
the required number of processes.

Consensus, Byzantine Faults, Fault-tolerant distributed algorithms, synchronous systems, realistic fault models

http://kathrin.dagstuhl.de/files/Materials/06/06371/06371.WidderJosef.Slides1.pdf