We study the security challenges faced by the mobile agent paradigm, where code travels and performs computations on remote hosts in an autonomous manner. We define universally composable security for mobile agent computation that is geared toward a complex networking environment where arbitrary protocol instances may be executing concurrently. Our definition provides security for all the participants in the mobile agent system: the originator as well as the hosts. Finally, under the assumption of a universally composable threshold cryptosystem, we present universally composable, multi-agent protocols with provable security against either static, semi-honest or static, malicious adversaries, according to our definition, where in the latter case we need to provide access to a common reference string.
Ke Xu, Stephen R. Tate