Section: New Results

Dishonest keys (Objective 2)

Participants : Hubert Comon-Lundh, Guillaume Scerri.

One of the main issues in the formal verification of the security protocols is the validity (and scope) of the formal model. Otherwise, it may happen that a protocol is proved and later someone finds an attack. This paradoxical situation may happen when the formal model used in the proof is too abstract.

A main stream of research therefore consists in proving full abstraction results (also called soundness): if the protocol is secure in the (symbolic) model, then an attack can only occur with negligible probability in a computational model. Such results have two main drawbacks: first they are very complicated, and have to be completed again and again for each combination of security primitives. Second, they require strong hypotheses on the primitives, some of which are not realistic. For instance, it is assumed that the attacker cannot forge his own keys (or that all keys come with their certificates, even for symmetric encryption keys).

Hubert Comon-Lundh, Véronique Cortier and Guillaume Scerri [31] propose an extension of the symbolic model, and prove it computationally sound, without this restriction on the dishonest keys.