Over the past decade, decentralized cryptocurrencies have received attention in industry and academia. Hardware wallets are dedicated devices that manage cryptocurrencies safely without entrusting cryptographic keys to a third party. Side-channel attacks have been widely studied in cryptanalysis and have already been proven threatening, but analysis on hardware wallets still needs to be researched. Although the previous work demonstrated several side-channel vulnerabilities, their attacks require a finely controlled environment or a learning phase of target devices’ physical properties before the attacks. This paper proposes a side-channel attack on hardware wallets extracting private keys. The proposed attack needs a single power trace measured when wallets process elliptic curve scalar multiplication with private keys. Our attack is reasonable since we do not damage the device under attack and do not target a specific device but an algorithm; it is widely applicable to wallets using that algorithm or analogous ones. It also presents the attack results conducted with three datasets: simulation, ChipWhisperer, and actual dataset collected from the Trezor Model One, the first and representative hardware wallets which comply with the de facto standard of hardware wallets.

With the development of 5G and network technology, the usage of IoT devices has become popular. Because most of these IoT devices can be controlled by an adversary away from the administrator, several security issues such as firmware dumping can arise. Firmware dumping is the cornerstone or goal of many types of hardware hacking. Therefore, many IoT device manufacturers adopt some protection mechanisms such as the restriction of hardware debuggers. However, several recent studies have shown that the operating instructions of an IoT device can be recovered through the profiling-based side-channel analysis. The Side-Channel-Based Disassembler (SCBD) refers to software that recovers instructions of the device only from the side-channel signal. The SCBD is powerful enough to defeat many firmware protection mechanisms. In this paper, we show how an adversary can build an instruction (opcode)-level disassembler using the power consumption signal of commercial microcontrollers (MCUs) such as the 8-bit ATxmega128 and 32-bit STM32F0. To implement the SCBD, we elaborately constructed the instruction template considering the pipeline of the target MCUs through instruction sequence analysis. Furthermore, we preprocessed the side-channel signals using the Continuous Wavelet Transform (CWT) for noise reduction and Kullback-Leibler Divergence (KLD) for instruction feature extraction. Our experimental results show that the machine-learning-based instruction disassembling models can recover the operating instructions with an accuracy of about 91.9% and 98.6% for the ATxmega128 and STM32F0, respectively. Furthermore, we achieved an accuracy of 77% and 96.5% in a cross-board validation.

Continuous Wavelet Transform (CWT) is rarely used in the field of side-channel analysis due to problems related to parameter (wavelet scale) selection; There is no way to find the optimal wavelet scale other than an exhaustive search, and the resulting spectrogram analysis can introduce significant analysis complexity. However, a well-scaled CWT can improve the signal-to-noise ratio of side-channel signals, which can lead to better attack performance. And our insights suggest that there is scope for CWT and deep learning approaches to be combined, which could help the models to train more effectively while overcoming the problems of CWT. In this context, we propose a novel feature extraction layer that combines a CWT with a Convolutional Neural Network (CNN). The proposed method can leverage neural network training to automatically adjust a wavelet scale, which is a critical parameter of CWT. Furthermore, the proposed method can lead to performance improvements by enabling a deep learning model to perform on-the-fly multi-frequency analysis without any pre-processing. By bringing the two approaches together, we were able to overcome the limitations of CWT and improve the performance of deep learning-based side-channel analysis. As an experimental result using open dataset ASCAD, a de facto standard in deep learning-based side-channel analysis, we confirmed that the proposed method could improve the performance by inserting the proposed layer into existing state-of-the-art deep learning models.