Enhancing Time-Frequency Analysis with Zero-Mean Preprocessing

Area of Science:

• Cryptography
• Embedded Systems Security
• Signal Processing

Background:

• Side-channel analysis poses a significant threat to IoT and embedded cryptosystems.
• Combined countermeasures like masking and desynchronization are cost-efficient but vulnerable to advanced attacks.
• Time-frequency analysis (TFA) was proposed to counter these defenses but shows limitations.

Purpose of the Study:

• Investigate factors affecting TFA performance against combined countermeasures.
• Mathematically analyze TFA preprocessing for second-order side-channel analysis.
• Propose and validate zero-mean preprocessing to enhance TFA effectiveness.

Main Methods:

• Mathematical analysis of TFA frequency information preprocessing.
• Experimental validation using two datasets of first-order masked Advanced Encryption Standard (AES) software implementations.
• Comparison of TFA performance with and without zero-mean preprocessing.

Main Results:

• Zero-mean preprocessing significantly enhances the performance of time-frequency analysis.
• The proposed preprocessing method improves TFA's ability to detect side-channel attacks.
• Experimental results confirm the enhanced or complementary performance of TFA with zero-mean preprocessing.

Conclusions:

• Zero-mean preprocessing is a crucial factor for improving TFA against masked cryptosystems.
• This technique offers a practical enhancement for physical security in embedded devices.
• The findings contribute to developing more robust defenses against sophisticated side-channel attacks.