Current systems providing anonymous interactive communication are based on networks of anonymity-providing relays called MIXes. An important issue with such systems is that a MIX is able to betray its users, and thus it is necessary to use several MIXes sequentially for each communication, which distributes the trust among them. This increases the complexity of the protocols as well as the latency. On the other side, such distributed systems are resilient and scalable, and they provide good enough performance for web browsing.

An ideal relay should be unable to betray its users (we will say in this case that the relay is trustable). In such a setting, using multiple relays to distribute trust is not necessary, which simplifies design and reduces costs. Superposed sending provides an approach to construct trustable relays, the DC-net relays. Straightforward usage of classic protocols leads to other approaches and recently we proposed a set of trustable relays, based on Private Information Retrieval protocols, that provide new alternatives.

Independently of their practical applications, these relays are interesting from a theoretic point of view. In this paper, we present a survey that gathers the different trustable relays we have been able to identify and gives a unified view of their construction.