18 June 2024 in IEEE Access
Side-channel analysis is one of the vulnerabilities in IoT’s cryptographic systems. To present reliable side-channel analysis results, it is crucial to collect considerable amounts of power consumption traces. Handling large datasets in this context can be highly inefficient regarding data transmission and storage. In various fields, the importance of compression techniques for efficient data storage has increased significantly. Compression techniques for various types of datasets are often designed with consideration for their data characteristics, much like JPEG for images. However, despite its relatively low compression rates, side-channel analysis researchers commonly use Deflate for data compression due to its simplicity and universality. In this paper, we propose a novel side-channel data compression technique using autoencoders, which offers higher compression rates than Deflate while maintaining decompression times and achieving a fast compression time. Furthermore, our model preserves the characteristics of the side-channel throughout the compression and decompression processes. To verify this, we conducted experiments comparing the original data with traces decompressed data using our technique through correlation power analysis. The results confirmed that they exhibit similar correlation coefficients and identical peak positions, demonstrating the preservation of essential data properties.
23 April 2026 in IACR Transactions on Cryptographic Hardware and Embedded Systems
The widespread deployment of AI applications on wearable and IoT devices has raised security concerns, including model stealing and personal information leakage. Model stealing attacks rely on issuing queries to the target model and analyzing the corresponding outputs. Side-channel analysis, traditionally used to attack cryptographic algorithms, can now facilitate the performance of model stealing attacks on neural networks (NN), by recovering their architectures and parameters. This exposure to model stealing presents risks such as intellectual property theft and adversarial attacks. Masking, a widely used countermeasure for side-channel attacks, has recently been adapted to NN to protect secret model information. However, existing masked NNs incur significant computational overhead or lead to accuracy degradation. In this work, we focus on reducing the main degradation in masking computational costs of non-linear operations specifically activation and argmax. As both operations hinge on sign bit extraction, we achieved an efficient realization using LuTs. Furthermore, we systematically identify and remove flawed masking gadgets that cause accuracy degradation and subsequently apply our gadgets to binarized neural network (BNN). As a result, our proposed BNN implementation securely maintains model accuracy while achieving a 42% performance improvement compared to state-of-the-art software-based masking methods [AWDF21].