Securing Agent Based Architectures

摘要

Agent based architectures provide significant flexibility and extensibility to software systems that attempt to model complex real world interactions between human users and functional agents. Such systems allow agents to be seamlessly published into the system providing services to human agent consumers. Securing agent based architectures in permissions based environments while still maintaining extensibility involves establishing a pathway of trust between the agent producer, container and consumer. This paper focuses on the final trust step, verifying the identity of an agent consumer in order to bound the capability of an agent by the capabilities of the agent consumer. We present an innovative application of zero knowledge proofs to inexpensively authenticate agents and grant them the restricted permissions of their consumer operator. Our scheme's theoretical foundation guarantees inexpensive detection of "rogue" agents and defends against replay attacks in environments where performance is critical.