A Security Enhancement and Proof for Authentication and Key Agreement (AKA)

Abstract

In this work, we consider Authentication and Key Agreement (AKA), a popular client-server Key Exchange (KE) protocol, commonly used in wireless standards (e.g., UMTS), and widely considered for new applications. We discuss natural potential usage scenarios for AKA, attract attention to subtle vulnerabilities, propose a simple and efficient AKA enhancement, and provide its formal proof of security.

The vulnerabilities arise due to the fact that AKA is not a secure KE in the standard cryptographic sense, since Client $\mathcal{C}$ does not contribute randomness to the session key. We argue that AKA remains secure in current deployments where $\mathcal{C}$ is an entity controlled by a single tamper-resistant User Identity Module (UIM). However, we also show that AKA is insecure if several Client’s devices/UIMs share his identity and key.

We show practical applicability and efficiency benefits of such multi-UIM scenarios. As our main contribution, we adapt AKA for this setting, with only the minimal changes, while adhering to AKA design goals, and preserving its advantages and features. Our protocol involves one extra PRFG evaluation and no extra messages. We formally prove security of the resulting protocol. We discuss how our security improvement allows simplification of some of AKA security heuristics, which may make our protocol more efficient and robust than AKA even for the current deployment scenarios.

Full version of this paper appears in ePrint archive [8].