Paper 2020/717

Fault Location Identification By Machine Learning

Anubhab Baksi, Santanu Sarkar, Akhilesh Siddhanti, Ravi Anand, and Anupam Chattopadhyay

Abstract

As the fault based analysis techniques are becoming more and more powerful, there is a need to streamline the existing tools for better accuracy and ease of use. In this regard, we propose a machine learning assisted tool that can be used in the context of a differential fault analysis. In particular, finding the exact fault location by analyzing the XORed output of a stream cipher/ stream cipher based design is somewhat non-trivial. Traditionally, Pearson's correlation coefficient is used for this purpose. We show that a machine learning method is more powerful than the existing correlation coefficient, aside from being simpler to implement. As a proof of concept, we take two variants of Grain-128a (namely a stream cipher, and a stream cipher with authentication), and demonstrate that machine learning can outperform correlation with the same training/testing data. Our analysis shows that the machine learning can be considered as a replacement for the correlation in the future research works.