Title: CONTROLLER MODULE

As noted above, the ability to control the communication of agents is generally not a feature of existing MAS architectures but at the same time a central requirement of the described approach for privacy-preserving information filtering. The required functionality cannot be realized based on regular agent services or components, because an agent on a platform is usually not allowed to interfere with the actions of other agents in any way. Therefore, additional infrastructure is added providing the required functionality to the MAS architecture itself, resulting in an agent environment with extended functionality and responsibilities.

Controlling the communication capabilities of an agent is realized by restricting via rules, in a manner similar to a firewall, but with the consent of the respective agent, its incoming and outgoing communication to specific platforms or agents on external platforms as well as other possible communication channels, such as the file system. Consent is required because otherwise the overall security would be compromised, as attackers could arbitrarily block various communication channels. The described approach does not require controlling the communication between agents on the same platform, and therefore this aspect is not addressed. Consequently, all rules addressing communication capabilities have to be enforced across entire platforms, because otherwise a controlled agent could just use a non-controlled agent on the same platform as a relay for communicating with agents residing on external platforms. Various agent services provide functionality for adding and revoking control of platforms, including functionality required in complex scenarios where controlled agents in turn control further platforms. The implementation of the actual control mechanism depends on the actual MAS architecture. In the implementation, methods provided via the Java Security Manager as part of the Java security model have been utilized. Thus, the supervisor agent is enabled to define custom security policies, thereby granting or denying other agents access to resources required for communication with other agents as well as communication in general, such as files or sockets for TCP/IP-based communication.