Home Projects Publications Presentations People News Activities About DCSL Internal
<< All Projects Privacy Preserving Communication

With the growing demand for adaptive and intelligent communication networks, content-based publish subscribe (CBPS) has gained significant attention in the research community over the last decade. Publish-subscribe in general is a communication technique whereby publishers "publish" messages and a set of subscribers receive these messages. This is more efficient than a publisher sending multiple point-to-point messages to each subscriber. Publish-subscribe offers a degree of decoupling between the publishers and the subscribers---a network of brokers together route the messages from a publisher to the correct set of subscribers. In CBPS systems, the subscriber defines a filter (a logical expression) on the content of a message, such as, a diabetic patient may be interested in availability of a drug named `Glucotrol' where the store zip code is either `47906' or `47907' and unit price is less than `$3'. Only messages (termed Notifications) matching the filter will be delivered to our hypothetical patient. This scenario is depicted in the figure below. Here the brokers execute sophisticated algorithms for matching messages with constraints on attribute values in the filter (such as sub-string, equality, inequality).


However, CBPS systems rely heavily on the integrity of brokers. Researchers have shown that achieving message confidentiality, integrity, and auditability in the presence of malicious brokers is a challenging assignment. Consider the following scenario: Our hypothetical patient Jane is infected with HIV. She wants to subscribe for drug availability and preventive care newsletters for her disease from an online health information exchange that uses CBPS for content delivery. Due to the sensitivity of her disease, Jane doesn't want the brokers to learn about her subscription. Similarly, Dr. Watson, a publisher in the health information exchange, doesn't want to divulge contents of his messages to the brokers. But normal functioning of CBPS requires that the brokers should inspect both pieces of information (notifications and subscriptions) to route messages. In this project, we solve this paradoxical problem of computing routing decisions based on encrypted notification and subscriptions, denoted hereafter as the Secure Routing Problem in CBPS. A solution to the orthogonal problem, that of computation and distribution of group keys in CBPS may be found elsewhere [Usenix Security 2001, Prakash et al.].

We make the important observation that routing in content-based publish-subscribe networks does not necessarily require inspection of the whole message. Instead, if a trusted publisher extracts the routing information from a message before encrypting it, then the problem reduces to hiding this information from malicious brokers. A level of indirection in naming of filters combined with separation of duty transforms the paradoxical problem to a practical one. In our solution approach, the publisher looks at the commonality of interests among subscribers and encodes the routing information in the form of a routing vector. The routing vector (RV) is added to the header of a message and it allows brokers to compute their receiver lists. We present two versions of our protocol---one where the RV is left unencrypted (termed the RV protocol), and the second where the RV is further encrypted to achieve both confidentiality and anonymity (termed Secure RV or SRV protocol). The collection of these two protocols is termed v-CAPS: Confidentiality and Anonymity preserving routing protocol for content-based Publish-Subscribe networks using routing vectors. Our simple approach eliminates the need for complex cryptographic operations, thereby, making it possible to incorporate the full generality of filters in baseline CBPS systems, with low computational overhead on the brokers. Our experimental results show that RV performs nearly as fast as a baseline CBPS in terms of latency. Achieving perfect anonymity (which we do through the SRV protocol), however, is significantly more costly and practical only for medium-sized networks.


Achieved Technical
Future Work
Code & Data
Funding Source
465 Northwestern Avenue, West Lafayette, IN 47907   |  dcsl@ecn.purdue.edu   |  +1 765 494 3510
Home |  Projects  |  Publications  |  Presentations  |  People
News  |  Activities |  About DCSL  |  Internal

Last Update: February 17, 2012 13:48 by GMHoward