I'd like to evaluate on Tor security metrics that are based on adversary models. Some adversary models to consider are
- An adversary that can compromise k relays.
- An adversary that can contribute b bandwidth.
- An adversary that can observe k Autonomous Systems. The security metrics for Tor based on these models that are relevant to Tor include
- The probability that the adversary observes the entry and exit traffic of a given connection.
- The probability distribution of the number of connections of a user for which the adversary observes entry and exit traffic (guard selection is particularly important here).
- The "linkability" of different connections.