The publish/subscribe communication model is increasingly considered for implementing middleware infrastructures for widely distributed applications. Scalability issues and routing algorithms of such systems have recently been the focus of intensive research. So far little attention has been given to security and management issues.

In current publish/subscribe systems, malicious publishers can very easily insert bogus notifications which may propagated to a large number of subscribers. Moreover, there is no method to control what notifications the subscribers are authorized to receive.

We describe a method to specify access control policy rules using expressions similar to subscription expressions. These policies define access rules for publish and subscribe functions and screening rules for notifications.