Fair Exchange in Strand Spaces

Abstract: Many cryptographic protocols are intended to coordinate state changes among principals. Exchange protocols coordinate delivery of new values to the participants, e.g. additions to the set of values they possess. An exchange protocol is fair if it ensures that delivery of new values is balanced: If one participant obtains a new possession via the protocol, then all other participants will, too. Fair exchange requires progress assumptions, unlike some other protocol properties. The strand space model is a framework for design and verification of cryptographic protocols. A strand is a local behavior of a single principal in a single session of a protocol. A bundle is a partially ordered global execution built from protocol strands and adversary activities. The strand space model needs two additions for fair exchange protocols. First, we regard the state as a multiset of facts, and we allow strands to cause changes in this state via multiset rewriting. Second, progress assumptions stipulate that some channels are resilient-and guaranteed to deliver messages-and some principals are assumed not to stop at certain critical steps. This method leads to proofs of correctness that cleanly separate protocol properties, such as authentication and confidentiality, from invariants governing state evolution. G. Wang's recent fair exchange protocol illustrates the approach.