Currently, untrusted third-party entities are increasingly involved in various stages of IC design and manufacturing, posing a significant threat to the reliability and security of SoCs due to the presence of hardware Trojans (HTs). In this paper, gate-level HT detection methods based on graph neural networks( GNNs) are established to overcome the defects of existing machine learning, which makes it difficult to characterize circuit connection relationships. We introduce harmonic centrality in the feature engineering of gate-level HT detection, which reflects the positional information of nodes and their adjacent nodes in the graph, thereby enhancing the quality of feature engineering. We use the golden section weight optimization algorithm to configure penalty weights to alleviate the problem of extreme data imbalance. In the SAED database, GraphSAGE-LSTM model obtained a TPR of 88.06% and an average F1 score of 90.95%. In the combined HT netlist of LEDA datasets, GraphSAGE-POOL model obtains a TPR of 88.50% and the best F1 score of 92.17%. In sequential HT netlist, GraphSAGE-LSTM model performs optimally, with a TPR of 98.25% and an average F1 score of 98.59%. Compared to existing detection models, the F1 score is enhanced by 8.86% and 2.48% on combined and sequential HT datasets, respectively.

The acoustic vibration method is an effective water pipe leak detection method, but when the detection pipeline is long, the surrounding environment is noisy or the leaks are small, the leakage signal will be completely submerged in noise, which leads to a huge error in leak location. Therefore, to improve the accuracy of localization, it is very necessary to extract the leak signal from the high-intensity noise. An adaptive leak signal extraction method based on the combination of robust local mean decomposition and singular value decomposition (RLMD-SVD) is proposed for water pipeline leak location. The original signal is decomposed into several product function (PF) components by the RLMD method, and then the effective PFs are selected to reconstruct the signals. Then, the Hankel matrix is constructed and decomposed by SVD, and further noise-reduced signals are obtained by selecting the effective rank order. In this paper, the principles and steps of the algorithm are elaborated, and the effectiveness is verified in simulations and experiments. In the simulations, two detection signals with signal-to-noise ratios (SNR) of -15 dB are improved by 10.26 dB and 10.37 dB after RLMD processing, and further improved by 1.76 dB and 1.79 dB after SVD processing. After RLMD-SVD processing, the peak of the cross-correlation function is more obvious and the time-delay is estimated accurately. In the full-size experiment, the location of the leakage can be found accurately after RLMD-SVD processing, and the results of RLMD-SVD processing are more accurate and reliable compared with no processing.

The localization accuracy of the water pipeline will be significantly impacted by noise for the acoustic-based method. Specifically, when the noise intensity exceeds that of the leakage signal, conventional noise reduction techniques tend to be ineffective. To address this problem, a hybrid method based on the gray wolf optimization algorithm to optimize variational modal decomposition and cross-correlation (GWO-VMD-CC) is proposed, which can adaptively extract effective leakage signals from strong noise. In this method, an objective function is innovatively established based on the correlation of leakage signals, and it is optimized by the efficient GWO so that the effective leakage signals are concentrated in one component of the VMD. The metric of the crest factor (CF) is defined to pick out this component to reduce the effect of strong noise. In addition, the 2σ principle is proposed to further improve the accuracy and reliability of the delay estimation. In simulation experiments, the effectiveness of the method is verified, and the results show that the noise intensity of the processed signals is significantly reduced, with the SNR of the two signals improved by 9.98 dB and 12.36 dB, respectively. The comparison experiments verify the superiority of this algorithm, and it has the most obvious noise reduction effect and the strongest anti-noise ability compared with the traditional method. Finally, the effectiveness and practicality of the method are further verified by prototype experiments, and the localization accuracy is significantly improved after being processed by the method.

In the field of leakage localization in water supply pipelines using acoustic-based methods, limited research has focused on addressing multiple leaks. The interaction among multiple leakage sources complicates the localization process. Targeting two leakage sources, this study presents a novel hybrid approach named Complementary Ensemble Empirical Modal Decomposition, Singular Value Decomposition, and Frequency-Domain Convolutive Blind Source Separation (CEEMD-SVD-FCBSS) for accurately localizing leakage points in water supply pipes. In this method, CEEMD and SVD are jointly proposed to adaptively extract the effective leakage signals, and FCBSS is innovatively proposed for estimating the time delay of each leakage source. In signal simulation, the effectiveness of the method is verified and the results show that after CEEMD-SVD noise reduction, the signal-to-noise ratios (SNRs) of two signals are improved significantly by 6.75 dB and 6.68 dB, respectively, and the time delays of the leak sources are found by the FCBSS. The comparative analysis highlights CEEMD-SVD’s superior noise reduction capabilities over conventional methods, while FCBSS effectively avoids the issue of large leaks masking smaller ones observed with the correlation method. A comprehensive full-size experiment further corroborates the method’s effectiveness and superiority in leakage localization in practical scenarios.

Linear prediction coding (LPC) is the core technology in speech processing, which has been successfully applied in speech recognition, synthesis and coding. LPC coefficient can well represent the speaker's vocal tract information, and is widely used in the field of speaker recognition. However, the epiphytic LPC residual is often ignored. This paper shows that the LPC residual contains information, which can reflect the characteristics of the speaker himself. We extracted new feature parameters (Second moment, third moment) from the LPC residual, combined with LPC coefficients, and input them into a speaker recognition system based on the GRU network. Cross entropy loss is used as a loss function to train the classifier. The experimental results show that the combined parameters effectively improve the system recognition rate by 5% relative to the LPC coefficients.