Double laser-faults based PFA on cryptographic circuits with algebraic analysis

Abstract

Cryptographic algorithms have been employed in a variety of fields as the primary method to protect information security. The security of a cryptographic algorithm is closely related to its operating environment and physical devices. Algebraic Persistent Fault Analysis (APFA) is a new fault analysis method for block ciphers proposed in CHES 2022, which utilizes the fault that persists in encryptions and introduces algebraic analysis in the fault analysis step. In the fault injection step, as the transistors of the integrated circuit are getting smaller and tighter, even high-precision devices may cause more than one fault per injection. However, more faults may lead to a more efficient attack in the fault analysis step. In this paper, APFA for double faults is proposed, which can deal with the double faults model and reduce the number of required ciphertexts. The practicality of our fault injection is validated by laser fault injection experiments on the SRAM embedded in an ATmega163L microcontroller. The effectiveness of our fault analysis is proven by successfully recovering the key of PRESENT-128 and AES-128. The number of ciphertexts needed for key recovery is reduced by 46% compared to PFA with a single fault.