Authenticated key agreement with key re-use in the short authenticated strings model

Abstract

Serge Vaudenay [20] introduced a notion of Message Authentication (MA) protocols in the Short Authenticated String (SAS) model. A SAS-MA protocol authenticates arbitrarily long messages sent over insecure channels as long as the sender and the receiver can additionally send a very short, e.g. 20 bit, authenticated message to each other. The main practical application of a SAS-MA protocol is Authenticated Key Agreement (AKA) in this communication model, i.e. SAS-AKA, which can be used for so-called "pairing" of wireless devices. Subsequent work [9,12,10] showed three-round SAS-AKA protocols. However, the Diffie-Hellman (DH) based SAS-AKA protocol of [10] requires choosing fresh DH exponents in each protocol instance, while the generic SAS-AKA construction given by [12] applies only to AKA protocols which have no shared state between protocol sessions. Therefore, both prior works exclude the most efficient, although not perfect-forward-secret, AKA protocols that re-use private keys (for encryption-based AKAs) or DH exponents (for DH-based AKAs) across multiple protocol sessions. In this paper, we propose a novel three-round encryption-based SAS-AKA protocol, using non-malleable commitments and CCA-secure encryption as tools, which we show secure (but without perfect-forward secrecy) if each player re-uses its private/public key across protocol sessions. The cost of this protocol is dominated by a single public key encryption for one party and a decryption for the other, assuming the Random Oracle Model (ROM). When implemented with RSA encryption the new SAS-AKA protocol is especially attractive if the two devices being paired have asymmetric computational power (e.g., a desktop and a keyboard). © 2010 Springer-Verlag Berlin Heidelberg.