Protecting co-operating mobile agents against malicious hosts

We propose a security model for open multi-agent systems. Given
a user-defined task T, we generate a set of mobile agents which
realize a common functionality that solves T. Those agents
co-operate with each other and build an autonomous community.
Using a scheme for secure distributed computations, this
community is able to perform secure computations without
requiring interaction with a trusted party. For this paper, we
have chosen Canetti's model for secure multi-party computations.
Unfortunately, the problems arising from the migration of agents
are not covered by this technique. We present an extended model
that offers a solutions to this. Thus, we yield guarantees for
confidentiality of secret data, detection of unauthorised code
and data changes, reestablishment of corrupted agents and
prevention from malicious routing.