Fault Tolerant Infective Countermeasure for AES

Abstract

Infective countermeasures have been a promising class of fault attack countermeasures. However, they have been subjected to several attacks owing to lack of formal proofs of security and improper implementations. In this paper, we first provide a formal information theoretic proof of security for one of the most recently proposed state of the art infective countermeasures against DFA, under the assumption that the adversary does not change the flow sequence or skip any instruction. Subsequently, we identify weaknesses in the infection mechanism of the countermeasure that could be exploited by attacks which change the flow sequence. Furthermore, we propose an augmented infective countermeasure scheme obtained by introducing suitable randomizations that reduce the success probabilities of such attacks. Finally, we develop a fault tolerant implementation of the countermeasure using the x86 instruction set to make any attacks which attempt to change the control flow of the algorithm via instruction skips practically infeasible. All the claims have been validated by supporting simulations and real-life experiments on a SASEBO-W platform. We also compare the fault tolerance provided by our proposed countermeasure scheme against that provided by the existing scheme.