Dissemination of anonymized streaming data

With the vision of the emergence of streaming data marketplaces, we study the problem of how to use a scalable dissemination infrastructure, composed by a number of brokers, to disseminate anonymized streaming data to a large number of clients. To satisfy the clients, who are trusted at different anonymity levels and have their own urgencies in requiring the data, we propose to deeply integrate the anonymization process into the dissemination infrastructure. More specifically, we extend the existing anonymization algorithms to derive the anonymity data from other anonymity data with different privacy constraints rather than only from the original microdata, in a technique which we call version derivation. With this flexibility, the anonymous data can be generated as needed, according to the available bandwidth, on the way from the data source to the end clients. Exploiting such new opportunities, we formulate the problem of dissemination planning which aims at minimizing the information loss of the disseminated data. Furthermore, we design two dissemination plan optimization strategies to solve the problem. The experimental study using both synthetic and real datasets verifies the effectiveness of our approach.