Computationally complete symbolic attacker and key exchange

Recently, Bana and Comon-Lundh introduced the notion of computationally complete symbolic attacker to deliver unconditional computational soundness to symbolic protocol verification. First we explain the relationship between their technique and Fitting's embedding of classical logic into S4. Then, based on predicates for "key usability", we provide an axiomatic system in their framework to handle secure encryption when keys are allowed to be sent. We examine both IND-CCA2 and KDM-CCA2 encryptions, both symmetric and asymmetric situations. For unforgeability, we consider INT-CTXT encryptions. This technique does not require the usual limitations of computational soundness such as the absence of dynamic corruption, the absence of key-cycles or unambiguous parsing of bit strings. In particular, if a key-cycle possibly corrupts CCA2 encryption, our technique delivers an attack. If it does not endanger security, the security proof goes through. We illustrate how our notions can be applied in protocol proofs.