[Resource Topic] 2023/893: Short paper: Diversity Methods for Laser Fault Injection to Improve Location Coverage

Welcome to the resource topic for 2023/893

Short paper: Diversity Methods for Laser Fault Injection to Improve Location Coverage

Authors: Marina Krček, Thomas Ordas, Stjepan Picek


In the first step of fault injection attacks, it is necessary to perform fault injections for target characterization to improve the chances of finding vulnerabilities that can be exploited in the second step. The second step is the attack, where the vulnerabilities are used to break the security. This work considers the parameter search on the laser fault injection parameters. While this work can also be adjusted to find exploitable faults, the objective here is to find as many faults as possible on the target device. The goal comes from a security evaluation perspective to perform a successful target characterization. Previous works propose several methods, such as the memetic algorithm or hyperparameter tuning techniques. However, we notice a problem concerning the convergence of such methods to one specific target region, which is beneficial for an attack where one parameter combination could be enough. Indeed, these search algorithms lead to many observed vulnerabilities, but most seem to come from the same area on the target, which could mean some crucial vulnerabilities are missed.
In this work, we propose considering the location coverage of the algorithms and offer two methods promoting diversity in tested parameter combinations to increase it while still finding many faults.We compare the grid memetic algorithm and evolution strategies to the performance of the memetic algorithm and random search. Our results show a benefit from introducing diversity to increase location coverage, but the overall number of vulnerabilities is decreased compared to the memetic algorithm. However, the number of unique locations with vulnerabilities is similar between the three evolutionary algorithms, with evolution strategies providing the most distant locations.

ePrint: https://eprint.iacr.org/2023/893

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 .