[Resource Topic] 2024/060: The Insecurity of Masked Comparisons: SCAs on ML-KEM’s FO-Transform

Welcome to the resource topic for 2024/060

The Insecurity of Masked Comparisons: SCAs on ML-KEM’s FO-Transform

Authors: Julius Hermelink, Kai-Chun Ning, Emanuele Strieder


NIST has released the draft standard for ML-KEM, and ML-KEM is actively used in several widely-distributed applications. Thus, the wide-spread use of ML-KEM in the embedded worlds has to be expected in the near future. This makes security against side-channel attacks a pressing matter.

Several side-channel attacks have previously been proposed, and one line of research have been attacks against the comparison step of the FO-transform. These attacks construct a decryption failure oracle using a side-channel. A recent work published at TCHES 2022 stresses the need for higher-order masked comparisons by presenting a horizontal attack and proposes a t-probing secure comparison operation. A subsequent work by D’Anvers, Van Beirendonck, and Verbauwhede improves upon the performance of several previous proposals.

In this work, we show that the latter masked comparison suffers from weakness similar to those identified in the former. We first propose an approximate template attack that requires only a very low number of traces for profiling and has an exceptionally high noise tolerance. We show that the profiling phase is not necessary and can be replaced by a vertical analysis of the distribution of certain points of interest without knowledge of the targeted values. Finally, we explain how a horizontal attack may construct a decryption failure oracle from a single trace.

We provide a leakage model of the targeted operations, which is based on the noisy Hamming weight model. Our evaluations are carried out on a physical device to stress the practicality of our attack. In addition, we simulate the attacks to determine the measurement noise levels that can be handled. We discuss the underlying causes for our attack, the difficulty of securing the Fujisaki-Okamoto transform in ML-KEM, and draw conclusion about the (in-)sufficiency of t-probing security in this context.

ePrint: https://eprint.iacr.org/2024/060

See all topics related to this paper.

Feel free to post resources that are related to this paper below.

Example resources include: implementations, explanation materials, talks, slides, links to previous discussions on other websites.

For more information, see the rules for Resource Topics .