On Computer-Assisted Verification of Timed Anonymity of Multi-Agent Systems

A multi-agent system is called anonymous if an adversary who can observe all the occurrences of events from the system cannot determine who is the "actor agent" of the events. Many agent systems such as e-voting or e-cash systems should guarantee the anonymity property because in such systems agents deal with privacy information. Also, the notion of anonymity can be found in a party game called Werewolf, also known as Mafia, the game is recently employed as a research benchmark in a AI community. For analyzing the anonymity of agents, we should deal with real-time systems. In this paper we show what formulae should be proven for the timed anonymity of multi-agent systems and we conduct a computer-assisted proof with an interactive theorem proving tool.