Secure Resource Access for Open Systems

In open distributed systems of mobile agents, where code from remote sites may run locally, protection of sensitive data and system resources is of paramount importance. We present a security-based typing system that provides such protection, using a mix of static and runtime typing; mobile agents are allowed access to local resources in accordance with security clearance. We formalize security violations as runtime errors and prove that, using our semantics, security violations cannot occur at “good” sites, i.e. sites under control of a particular administrative domain.