Drac is a peer-to-peer communication system designed to make IM and VoIP traffic anonymous and unobservable. It achieves this goal by exposing the social connections of the users who make up the nodes of the peer-to-peer network.
"Drac gives away the identity of a user's friends to guarantee the unobservability of actual calls, while still providing anonymity when talking to trusted third parties," explains a paper on the proposed technology.
The paper, "Drac: An Architecture for Anonymous Low-Volume Communications," will be presented at the The 10th Privacy Enhancing Technologies Symposium (PETS 2010) in Berlin, Germany in July.
The paper was written by George Danezis, from Microsoft Research in Cambridge, Claudia Diaz and Carmela Troncoso at The Catholic University of Leuven (K.U. Leuven), and Ben Laurie, a computer security researcher.
The authors note that while anonymous online communication may conceal the content of conversations, information about the network addressing, the timing of the messages, and the volume of traffic often reveals as much as the hidden correspondence.
Drac aims to preserve anonymity while also thwarting traffic analysis by using a peer-to-peer relay architecture that routes data through social networking connections.
The Drac system envisions a network of friends who have a strong trust relationship and who share cryptographic keys to maintain secure communication links. Their social networking connections become the possible network data paths for Drac messaging.
In addition, Drac allows users to communicate with contacts outside their network. But while social network connections in Drac are public, contacts sending or receiving messages to or from the network are concealed.
The authors argue that relaying messages over a friend-of-a-friend network makes denial-of-service and related attacks less likely, does not require a central server or trust infrastructure, and avoids network discovery and random sampling attacks that affect other peer-to-peer systems.
The system is not without drawbacks: network paths over social networks tend to be longer, the authors observe, and knowledge of participants' social network connections may be undesirable in some contexts.
The Drac system has been implemented in a limited software simulator, the code for which the authors intend to make available on request. It's merely proof-of-concept code and Drac will have to be refined further before it emerges as a functional product.