A Privacy-Enhancing Content-Based Publish/Subscribe System Using Scalar Product Preserving Transformations
Users of content-based publish/subscribe systems (CBPS) are interested in receiving data items with values that satisfy certain conditions. Each user submits a list of subscription specifications to a broker, which routes data items from publishers to users. When a broker receives a notification that contains a value from a publisher, it forwards it only to the subscribers whose requests match the value. However, in many applications, the data published are confidential, and their contents must not be revealed to brokers. Furthermore, a user’s subscription may contain sensitive information that must be protected from brokers. Therefore, a difficult challenge arises: how to route publisher data to the appropriate subscribers without the intermediate brokers learning the plain text values of the notifications and subscriptions. To that extent, brokers must be able to perform operations on top of the encrypted contents of subscriptions and notifications. Such operations may be as simple as equality match, but often require more complex operations such as determining inclusion of data in a value interval. Previous work attempted to solve this problem by using one-way data mappings or specialized encryption functions that allow evaluation of conditions on ciphertexts. However, such operations are computationally expensive, and the resulting CBPS lack scalability. As fast dissemination is an important requirement in many applications, we focus on a new data transformation method called Asymmetric Scalar-product Preserving Encryption (ASPE) . We devise methods that build upon ASPE to support private evaluation of several types of conditions. We also suggest techniques for secure aggregation of notifications, supporting functions such as sum, minimum, maximum and count. Our experimental evaluation shows that ASPE-based CBPS incurs 65% less overhead for exact-match filtering and 50% less overhead for range filtering compared to the state-of-the-art.
Publish Subscribe Systems, Privacy, Confidentiality, Security
Date of this Version