Techniques for content subscription anonymity with distributed brokers

When issuing a one-shot or continous content-based subscription, there is an inherent tradeoff between the privacy of the subscriber and the accuracy of the matching notifications. The former can be described in terms of how well the exposed information uniquely characterized the subscriber, and the latter how well the returned data items match the subscriber's real interest. In this paper, we define the partial order basked k-filter anonymity, which generalizes k-location anonymity by allowing the generalization of filters using the subsumption or covering relation.