Authenticated Confidential Channel Establishment and the Security of TLS-DHE

Abstract

Transport Layer Security (TLS) is the most important cryptographic protocol in use today. However, finding a cryptographic security proof for the complete, unaltered protocol has proven to be a challenging task. We give the first such proof in the standard model for the core cryptographic protocol underlying TLS cipher suites based on ephemeral Diffie–Hellman key exchange (TLS-DHE). This includes the cipher suite TLS_DHE_DSS_WITH_3DES_EDE_CBC_SHA, which is mandatory in TLS 1.0 and TLS 1.1. It is impossible to prove the TLS Handshake secure in the classical security models of Bellare–Rogaway and Canetti–Krawczyk. The reason for this is that the final Finished messages of the TLS Handshake are encrypted with the session key, which provides an opportunity to distinguish real keys from random values. Therefore we start with proving the security of a truncated version of the TLS Handshake protocol, which has also been considered in previous work on TLS, and give the first proof of this variant in the standard model. Then we define the new notion of authenticated and confidential channel establishment (ACCE), which allows the monolithic analysis of protocols for which a modular security proof is not possible. We show that the combination of the TLS-DHE Handshake protocol and the TLS Record Layer encryption is secure in this model. Since the conference publication of this paper, the notion of ACCE has found many further applications, for example to the analysis of further TLS cipher suites (Krawczyk et al., Crypto 2013; Li et al., PKC 2014), advanced mechanisms like secure renegotiation of TLS session keys (Giesen et al., CCS 2013), and other practical protocols like EMV channel establishment (Brzuska et al., CCS 2013), SSH (Bergsma et al., CCS 2014), and QUIC (Lychev et al., S&P 2015).