A new correlation frequency analysis of the side channel

Abstract

Security in embedded computing systems is now an important concern for a diverse set of applications. However, the embedded hardware implementation may unintentionally leak information, through its electromagnetic emanations or current draw, which may lead to the revelation of secrets used in the cryptographic computations being performed. This paper presents an attack methodology and an empirical study, based on Correlation Analysis in the Frequency domain (CAF) with pre-characterization of the embedded system. Unlike previous research this analysis exploits the fact that a few frequencies are more likely to leak computing information, and are independent of the system clock (rather a function of the technology). Results indicate that the secret key can be reliably extracted from both hardware and software implementations of AES. The analysis presented is additionally tolerant to trace misalignments and has been tested with real power and electromagnetic (EM) traces used to extract 8-bit keys and full 128-bit keys. This research is important for providing more secure cryptographic computations necessary in many embedded systems.