Hardware Trojan detection method has been given particular attentions by hardware security researchers since the failure of Syrian radars in 2007. Electromagnetic side-channel analysis is a promising method which is widely used in practice due to its advantage of efficiency, non-touch and high accuracy. In this brief, we propose a novel method using electromagnetic side-channel signal for hardware Trojan detection by transfer learning. Firstly, time-frequency information of electromagnetic signal is extracted by continuous wavelet transform to take full advantage of useful information in time domain and frequency domain. Then, time-frequency information is fed to transfer learning network to get the classification result. In order to further improve the result, we use transfer learning to further extract the key features in time-frequency information. Finally, the key features are classified by the support vector machine to improve the accuracy. The experiment is conducted on a stand FPGA board and advanced encryption standard circuit is used as the benchmark circuit. Experimental results show that our methods can improve the result efficiently.

This paper focuses on electromagnetic information security in communication systems. Classical correlation electromagnetic analysis (CEMA) is known as a powerful way to recover the cryptographic algorithm’s key. In the classical method, only one byte of the key is used while the other bytes are considered as noise, which not only reduces the efficiency but also is a waste of information. In order to take full advantage of useful information, multiple bytes of the key are used. We transform the key into a multidimensional form, and each byte of the key is considered as a dimension. The problem of the right key searching is transformed into the problem of optimizing correlation coefficients of key candidates. The particle swarm optimization (PSO) algorithm is particularly more suited to solve the optimization problems with high dimension and complex structure. In this paper, we applied the PSO algorithm into CEMA to solve multidimensional problems, and we also add a mutation operator to the optimization algorithm to improve the result. Here, we have proposed a multibyte correlation electromagnetic analysis based on particle swarm optimization. We verified our method on a universal test board that is designed for research and development on hardware security. We implemented the Advanced Encryption Standard (AES) cryptographic algorithm on the test board. Experimental results have shown that our method outperforms the classical method; it achieves approximately 13.72% improvement for the corresponding case.

Correlation electromagnetic analysis (CEMA) is a method prevalent in side-channel analysis of cryptographic devices. Its success mostly depends on the quality of electromagnetic signals acquired from the devices. In the past, only one byte of the key was analyzed and other bytes were regarded as noise. Apparently, other bytes’ useful information was wasted, which may increase the difficulty of recovering the key. Multi-objective optimization is a good way to solve the problem of a single byte of the key. In this work, we applied multi-objective optimization to correlation electromagnetic analysis taking all bytes of the key into consideration. Combining the advantages of multi-objective optimization and genetic algorithm, we put forward a novel multi-objective electromagnetic analysis based on a genetic algorithm to take full advantage of information when recovering the key. Experiments with an Advanced Encryption Standard (AES) cryptographic algorithm on a Sakura-G board demonstrate the efficiency of our method in practice. The experimental results show that our method reduces the number of traces required in correlation electromagnetic analysis. It achieved approximately 42.72% improvement for the corresponding case compared with CEMA.

Embedded encryption devices and smart sensors are vulnerable to physical attacks. Due to the continuous shrinking of chip size, laser injection, particle radiation and electromagnetic transient injection are possible methods that introduce transient multiple faults. In the fault analysis stage, the adversary is unclear about the actual number of faults injected. Typically, the single-nibble fault analysis encounters difficulties. Therefore, in this paper, we propose novel ciphertext-only impossible differentials that can analyze the number of random faults to six nibbles. We use the impossible differentials to exclude the secret key that definitely does not exist, and then gradually obtain the unique secret key through inverse difference equations. Using software simulation, we conducted 32,000 random multiple fault attacks on Midori. The experiments were carried out to verify the theoretical model of multiple fault attacks. We obtain the relationship between fault injection and information content. To reduce the number of fault attacks, we further optimized the fault attack method. The secret key can be obtained at least 11 times. The proposed ciphertext-only impossible differential analysis provides an effective method for random multiple faults analysis, which would be helpful for improving the security of block ciphers.

The application of machine learning techniques in side-channel analysis has recently received increased attention. Finding the best hyperparameters to achieve optimal performance for machine learning models in side-channel analysis is still a challenging endeavor. In order to solve the problem, we present an efficient ensemble framework designed to support profiled side-channel analysis for attacking cryptographic devices with countermeasures. Our proposed framework can partially mitigate the impact of traditional countermeasures employed in cryptographic devices. Additionally, we introduce a novel voting method called elite voting, which leverages candidate keys with higher probabilities to recover the secret key and adjusts the voting weights for better candidate keys. Experimental results illustrate that our proposed framework can effectively recover the right key from cryptographic devices with countermeasures through multiple experiments. It enhances the signal-to-noise ratio of traces and successfully recovers the right key across various datasets. Furthermore, when compared to traditional methods, our elite voting method further enhances the performance of ensemble learning by reducing the number of traces needed to recover the secret key. It exhibits superior performance compared to other ensemble methods, as it can reduce the minimum required number of traces significantly.

Horizontal analysis is a widely adopted method in side-channel analysis, particularly for public-key cryptosystems (PKCs), where attackers aim to recover the key from a single trace. Current methods rely on trace segmentation, dimensionality reduction, and classification, but high noise and poor feature preservation hinder accuracy. Noise blurs cryptographic operation segment boundaries, and existing dimensionality reduction techniques fail to maintain the inherent distribution of trace points in a high-dimensional space. As a result, secret information recovery based on clustering remains inaccurate. This article proposes an automated horizontal analysis framework named uniform manifold approximation and projection (UMAP)-HC to improve secret information recovery accuracy. The framework uses a sliding segmentation method to locate cryptographic operations in noisy traces with blurred segment boundaries. It leverages UMAP for feature preservation and hierarchical clustering for secret information recovery. Experimental results on four open access public-key algorithm power trace datasets, an SM2 power trace collected from a smart card, and an ECC power trace with dummy operation countermeasures demonstrate that UMAP-HC effectively classifies cryptographic operations, accurately locates operation segments, and recovers secret key. It achieves up to 100% recovery accuracy, surpassing previous methods by 40%–70%, with normalized mutual information (NMI) reaching 1, an improvement of 0.02–0.99 over existing approaches.