An automated framework for exploitable fault identification in block ciphers

Abstract

Faults have been practically exploited on several occasions to compromise the security of mathematically robust cryptosystems at the implementation level. However, not every possible fault within a cryptosystem is exploitable for fault attack. Comprehensive knowledge about the exploitable part of the fault space is thus imperative for both the algorithm designer and the implementer in order to invent precise countermeasures and robust algorithms. This paper addresses the problem of exploitable fault characterization in the context of differential fault analysis attacks on block ciphers. A generic and automated framework has been proposed, which can determine the exploitability of fault instances from any given block cipher in a fast and scalable manner. Such automation is supposed to work as the core engine for analysing the fault spaces, which are, in general, difficult to characterize with manual effort due to their formidable size and the complex structural features of the ciphers. Our framework significantly outperforms another recently proposed one as reported by Khanna et.al. (in: DAC, ACM, pp. 1–6, 2017), in terms of attack class coverage and automation effort. Evaluation of the framework on AES and PRESENT establishes the efficacy of it as a potential tool for exploitable fault analysis.