The "Mobile Adversary" Paradigm in Distributed Computation and Systems

The notion of `mobile adversary,' where the opponent can capture parties in a multi-party protocol dynamically, as long as at any given point in time its capturing capability is limited by a bound on number of parties (processors) it can control, has been suggested as an extensions of the traditionally static adversary. The motivation for this adversary was a result of a few issues: First, it was originated in systems corruption phenomena like virus injection into a computer network (such as the Internet), where viruses are spread but also detected and eliminated at network computers. Secondly, it is natural to assume that for managed networks, there will be efforts to recover failed processors, so the adversary control may cease to exist in a node locally while its control of other nodes continues. Thirdly, the notion of self-stabilization implies that a system tries to get rid of failures and the notion represents a strong fault-tolerance aspect, and should be extended (i.e., augmented by assuming bound on faults and extending initial benign fault to adversarial perpetual ones). As a result, the mobile adversary notion and a methodology for coping with it (proactive fault-tolerance, or proactive security in security oriented protocols) and its perpetual self-healing character, have been suggested. This work will review the development and influence of this notion on fault-tolerant secure distributed computing.