With Great Power Come Great Side Channels: Statistical Timing Side-Channel Analyses with Bounded Type-1 Errors


Martin Dunsche, Marcel Maehren, and Nurullah Erinola, Ruhr University Bochum; Robert Merget, Technology Innovation Institute; Nicolai Bissantz, Ruhr University Bochum; Juraj Somorovsky, Paderborn University; Jörg Schwenk, Ruhr University Bochum


Constant-time implementations are essential to guarantee the security of secret-key operations. According to Jancar et al. [42], most cryptographic developers do not use statistical tests to evaluate their implementations for timing side-channel vulnerabilities. One of the main reasons is their high unreliability due to potential false positives caused by noisy data. In this work, we address this issue and present an improved statistical evaluation methodology with a controlled type-1 error (α) that restricts false positives independently of the noise distribution. Simultaneously, we guarantee statistical power with increasing sample size. With the bounded type-1 error, the user can perform trade-offs between false positives and the size of the side channels they wish to detect. We achieve this by employing an empirical bootstrap that creates a decision rule based on the measured data.

We implement this approach in an open-source tool called RTLF and compare it with three different competitors: Mona, dudect, and tlsfuzzer. We further compare our results to the t-test, a commonly used statistical test for side-channel analysis. To show the applicability of our tool in real cryptographic network scenarios, we performed a quantitative analysis with local timing measurements for CBC Padding Oracle attacks, Bleichenbacher's attack, and the Lucky13 attack in 823 available versions of eleven TLS libraries. Additionally, we performed a qualitative analysis of the most recent version ofeach library. We find that most libraries were long-time vulnerable to at least one of the considered attacks, with side channels big enough likely to be exploitable in a LAN setting. Through the qualitative analysis based on the results of RTLF, we identified seven vulnerabilities in recent versions.

Open Access Media

USENIX is committed to Open Access to the research presented at our events. Papers and proceedings are freely available to everyone once the event begins. Any video, audio, and/or slides that are posted after the event are also free and open to everyone. Support USENIX and our commitment to Open Access.