Demarcation of Security in Authentication Protocols

Security analysis of communication protocols is a slippery business; many “secure” protocols later turn out to be insecure. Among many, two complains are more frequent: inadequate definition of security and unstated assumptions in the security model. In our experience, one principal cause for such state of affairs is an apparent overlap of security and correctness, which may lead to many sloppy security definitions and security models.

Although there is no inherent need to separate security and correctness requirements, practically, such separation is significant. It makes security analysis easier, and enables us to define security goals with a fine granularity. We present one such separation, by introducing the notion of binding sequence as a security primitive.

A binding sequence, roughly speaking, is the only required security property of an authentication protocol. All other authentication goals, the correctness requirements, can be derived from the binding sequence.