Mobile Byzantine Agreement on Arbitrary Network

Brief Announcement: Optimal Self-stabilizing Mobile Byzantine-Tolerant Regular Register with Bounded Timestamps

“…As in the case of round-based MBF models [3,5,11,16,23], we assume that any process has access to a tamper-proof memory storing the correct protocol code.…”

“…Failures are represented by Byzantine agents that are managed by an omniscient adversary that "moves" them from a host process to another and when an agent is in some process it is able to corrupt it in an unforeseen manner. Models investigated so far in the context of mobile Byzantine failures consider mostly round-based computations, and can be classified according to Byzantine mobility constraints: (i) constrained mobility [11] agents may only move from one host to another when protocol messages are sent (similarly to how viruses would propagate), while (ii) unconstrained mobility [3,5,16,21,22,23] agents may move independently of protocol messages. In the case of unconstrained mobility, several variants were investigated [3,5,16,21,22,23]: Reischuk [22] considers that malicious agents are stationary for a given period of time, Ostrovsky and Yung [21] introduce the notion of mobile viruses and define the adversary as an entity that can inject and distribute faults; finally, Garay [16], and more recently Banu et al [3], and Sasaki et al [23] or Bonnet et al [5] consider that processes execute synchronous rounds composed of three phases: send, receive, and compute.…”

“…Hence the set of faulty processes at any given time has a bounded size, yet its membership may evolve from one round to the next. The main difference between the aforementioned four works [3,5,16,23] lies in the knowledge that hosts have about their previous infection by a Byzantine agent. In Garay's model [16], a host is able to detect its own infection after the Byzantine agent left it.…”

“…In Garay's model [16], a host is able to detect its own infection after the Byzantine agent left it. Sasaki et al [23] investigate a model where hosts cannot detect when Byzantine agents leave. Finally, Bonnet et al [5] considers an intermediate setting where cured hosts remain in control on the messages they send (in particular, they send the same message to all destinations, and they do not send obviously fake information, e.g.…”

“…To the best of our knowledge, the problem of tolerating both arbitrary transient faults and mobile Byzantine faults has been considered recently only in roundbased synchronous systems [6]. The authors propose optimal unbounded selfstabilizing atomic register implementations for round-based synchronous systems under the four Mobile Byzantine models described in [3,5,16,23].…”