Hydra: practical metadata security for contact discovery, messaging, and voice calls

Protecting communications’ metadata can be as important as protecting their content, i.e., recognizing someone contacting a medical service may already allow to infer sensitive information. There are numerous proposals to implement anonymous communications, yet none provides it in a strong (but feasible) threat model in an efficient way. We propose Hydra, an anonymity system that is able to efficiently provide metadata security for a wide variety of applications. Main idea is to use latency-aware, padded, and onion-encrypted circuits even for connectionless applications. This allows to implement strong metadata security for contact discovery and text-based messages with relatively low latency. Furthermore, circuits can be upgraded to support voice calls, real-time chat sessions, and file transfers - with slightly reduced anonymity in presence of global observers. We evaluate Hydra using an analytical model as well as call simulations. Compared to other systems for text-based messaging, Hydra is able to decrease end-to-end latencies by an order of magnitude without degrading anonymity. Using a dataset generated by performing latency measurements in the Tor network, we further show that Hydra is able to support anonymous voice calls with acceptable quality of service in real scenarios. A first prototype of Hydra is published as open source.