Reversible data hiding techniques have garnered increasing scholarly attention for their capacity to generate high-fidelity watermarked images while enabling lossless restoration of the original images. Pixel value ordering (PVO) stands as a pivotal reversible data embedding approach that adeptly leverages pixel sequencing within an image block to facilitate high-fidelity image embedding and reversible retrieval. The efficacy of PVO-based methodologies hinges on the block size and classification thresholds. Nonetheless, many techniques ascertain the appropriate parameters through exhaustive searches within confined search spaces, demanding significant computational time. This paper introduces a novel pixel value ordering method based on differential evolution (PVO-DE) to address these challenges more effectively. PVO-DE dynamically identifies the optimal or suitable block size and threshold based on distinct images and embedding quantities, thereby optimizing image utilization and reducing time expenditures. Furthermore, a complexity algorithm, fused with the differential evolution algorithm, is proposed to harness the strengths of diverse complexity algorithms and pinpoint the algorithm most conducive to the current image for optimal performance. Results demonstrate a reduction in computational costs for determining the optimal parameter set, alongside enhanced performance of marked images compared to prior approaches.

Side-Channel Attacks (SCAs) now require more side-channel traces for successful execution, which places more stringent requirements on the storage capacity and computational ability of the devices on which SCAs are based. To reduce the storage and computational pressure on the local device where SCAs are performed on collected leakage traces from an attacking device, this paper proposes a secure cloud outsourcing protocol to perform Principal Component Analysis (PCA) dimensionality reduction on the side-channel traces. Secure cloud outsourcing is applied for the computationally intensive matrix multiplication and eigenvalue decomposition of the PCA process. The proposed protocol has been proven to balance privacy, efficiency, and correctness. Through experiments on CW and Grizzly datasets, it shows that 1) Correlation Power Analysis (CPA) with PCA effectively mitigates noise, improving the probability of a successful CPA; 2) Cloud-based PCA significantly reduces the computational complexity of local devices; 3) Template Attacks (TAs) are performed on leakage trace data using cloud-based PCA, client-based PCA, Linear Discriminant Analysis (LDA) and Independent Component Analysis (ICA). The attack results of cloud-based and client-based PCA are basically identical, and they achieve lower Guessing-Entropy (GE) than ICA. Both theoretical analysis and experimental results demonstrate the feasibility and advantages of this protocol.

Deploying Deep Neural Network (DNN) models of varying capabilities to support collaborative inference for User Equipment (UE) tasks is becoming increasingly common in Mobile Edge Computing (MEC) systems. In this paper, we aim to minimize energy consumption in a dynamic Non-Orthogonal Multiple Access (NOMA)-based MEC system assisted by a Simultaneously Transmitting and Reflecting Reconfigurable Intelligent Surface (STAR-RIS) under complex Non-Line-of-Sight (NLoS) conditions, while meeting latency requirements and preserving inference accuracy for UEs. We formulate the problem as a non-convex optimization and address it using a Decomposition-Based Twin Delayed Deep Deterministic Policy Gradient (DB-TD3) approach. The problem is decomposed into two subproblems: 1) computation resource allocation and power optimization, and 2) optimization of the offloading ratio, time fractions allocated for reflection and transmission, phase shift, and transmission time. For the first subproblem, we derive optimal CPU frequencies and transmit power allocation through theoretical analysis. For the second subproblem, the offloading ratio, time fractions allocated for reflection and transmission, phase shift, and transmission time are optimized using the TD3 algorithm. Experimental results demonstrate that the DB-TD3 method significantly improves system efficiency and reduces average energy consumption by 59.3% compared to baseline algorithms. Furthermore, the proposed NOMA with STAR-RIS scheme outperforms other offloading methods, achieving an average energy reduction of 41.3%.

Side-channel attacks (SCAs) serve as a fundamental tool for evaluating the implementation security of cryptographic devices. In real-world acquisition scenarios, however, power consumption traces are frequently degraded by device clock jitter and external noise interference, resulting in pronounced temporal misalignment and signal distortion. As a result, the efficiency and stability of attack convergence are seriously restricted. To address this, we propose a correlation power analysis (CPA) framework that integrates successive variational mode decomposition (SVMD) and peak-anchor-based alignment (PA-CPA). Firstly, the method uses SVMD to adaptively decompose and reconstruct the original power consumption traces, effectively suppressing random interference and preserving leakage-relevant features. A robust anchor point sequence is then constructed and global linear resampling and local dynamic time warping (DTW) are combined to realise the segmental fine alignment of the power consumption traces. This improves feature synchronisation and alignment accuracy. Experimental results demonstrate that the proposed method achieves higher attack success rates and faster convergence.

ABSTRACT Epilepsy is a complex chronic brain disorder that can be identified by observing brain signals. In general, the electroencephalogram (EEG) can be used to detect these brain signals. In order to produce a high‐quality model, data from numerous patients can be gathered on a central server. However, sending the patient's raw data to the central computer may lead to privacy leakage. To address this problem, this work uses federated learning and differential privacy to train the model jointly. Furthermore, the epilepsy data is unbalanced as seizure only happens for a minority of time in one day, which influences the performance of the model. Thus, this work also uses label‐distribution‐aware‐margin (LDAM) loss to solve this issue. This work is evaluated in intracranial EEG datasets, which consist of two dogs' EEG records. The global model trained jointly with LDAM loss can achieve an accuracy of 96.95%, a sensitivity of 78.9%, a specificity of 96.145%, an F1 score of 70.435%, and a geometric mean of 87.785%. Compared with the other works, the accuracy has improved by about ˜9.31%, while the specificity and the geometric mean have also improved by about ˜10.75% and ˜1.8%, respectively.

Abstract Chaotic systems play an indispensable role in the fields of cryptography and information security. Sine-Transform-Based Chaotic System (STBCS) can address the shortcomings of low complexity and limited chaotic behaviour of classical chaos systems. In this paper, a compact hardware STBCS is proposed and developed on the FPGA device by using the Stochastic Computation (SC) technique. The traditional arithmetic operations are replaced by the SC and finite state machines design. The structure of STBCS is optimised, where the disturbance method is employed to improve the chaotic behaviours and also taking the SC method into account for implementation. The hardware performance of the proposed design is verified via various tests of the chaotic system and corresponding random number generator. Experimental results show that the utilisation of the hardware resources is reduced especially the DSP components compared to the traditional design methods. This provides an efficient design for the random generator of the alternative cryptosystems.

Random Boolean Network (RBN) is a type of regulatory network in which the nodes have Boolean values representing their states. The robustness of RBNs against perturbations is a crucial characteristic, and there has been a growing interest in enhancing the network's robustness. In this study, a biologically inspired epigenetic regulation method is proposed to enhance the robustness of the RBNs. A frequency encoding method based on pulse counting is employed to encode the node states within a sliding time window, thereby improving the form of epigenetic regulation. To verify the performance of this method, an antifragility indicator is adopted to measure the robustness of RBNs and yeast cell networks at different scales. The experimental results demonstrate that the networks with epigenetic regulation exhibit excellent robustness, even in the presence of large-scale networks and severe perturbations. This approach provides a new perspective and idea for designing robust RBNs and discrete networks.

Federated Learning (FL) represents a promising distributed learning paradigm that enables model training without centralizing users' sensitive data. However, FL faces several practical challenges, such as communication overhead, convergence rates, robustness, and overall performance efficacy. A paramount research objective within FL is to achieve rapid convergence without compromising privacy. Regrettably, the existing body of research on this subject is scant, with most studies concentrating solely on privacy safeguards or the optimization of adaptive learning algorithms. To address this gap, this work introduces an adaptive gradient descent algorithm underpinned by Differential Privacy (DP). The approach synergizes the TensorFlow federated learning framework with a suite of techniques, including relaxed DP, subsampling for privacy amplification, and the shuffle model, to expedite convergence while bolstering privacy protection. The empirical outcomes on the MNIST dataset show that the FAdagrad algorithm accomplishes a test accuracy of 82.29% within just 12 iterations, all the while maintaining a tighter privacy control (with a privacy budget of 0.5). In comparison to the DP-FL, Simple, and Topk algorithms that employ conventional differential privacy techniques and attain peak test accuracies of 52.17%, 71.09%, and 72.67% respectively under analogous conditions, the FAdagrad algorithm introduced in this paper markedly improves the key performance indicators of the model.

Video-based behavior recognition is a challenging research topic. The three dimensional convolution neural network (3D CNN) is effectively adopted to capture features from videos directly. 3D CNN is extended by two-dimensional convolution neural network, in which a time dimension is added. 3D CNN is better than two-dimensional convolution network in expressing effective motion information, and it has certain advantages. In order to make better use of the valuable features extracted from the original video information, only stacked RGB frame data sets can be used as the input of network. Ensembling 3D CNN framework for video recognition is proposed in the paper. Firstly, the pre-training model of Sports-1M is initialized firstly, and a 3D convolution neural network based on multi-level feature fusion is constructed. . The final high-dimensional feature combination is obtained by fusing multiple convolution features. Then 3D convolutional neural network based on ensemble learning is proposed to increase motion information, enrich motion features and enhance the robustness of single feature representation. Three incomplete training data sets are obtained by Bagging algorithm. To get different networks, three data sets are employed to train three 3D convolution neural networks respectively, and the output of the three networks is integrated. The output features of the three networks are input into the SVM classifier through the Stacking algorithm and the final results are obtained. The integration effects of different ensemble methods are compared. The experimental results show that the method of this work can improve recognition accuracy on UCF-101 data set effectively.

Recently deep learning has revolutionized the field of machine learning, for pattern recognition in particular. A deep neural network (DNN) requires a large number of labeled training samples, and the recognition accuracy is truly impressive, sometimes outperforming humans. Neurons in an artificial neural network (ANN) are modeled by a single, static, continuous-valued activation function. However, biological neurons employ discrete spikes to compute and transmit information. The information can be encoded by the spike times, spike rates, and spike phase, etc. As the third generation artificial neural networks, spiking neural networks (SNNs) are more closely mimic natural neural networks. SNNs can achieve the same goals as ANNs, and it has the ability to build a large-scale network structure (i.e. deep spiking neural network) to accomplish complex tasks. In this paper, a state-of-the-art manner, differential evolving spiking neural network (DESNN), is proposed for pattern classification. The XOR task, Iris data, and hand-written digits classification task on MNIST are used to validate the proposed training method. The experimental results show that the algorithm used in this work applies the fewer neurons and it is effective for pattern classification tasks.

Forest fires is one of the main causes of environmental degradation and its detection and forecasting is challenging. A novel method of forest fire detection based on spiking neural networks is proposed in this paper. Data obtained from controlled experiments are used as input training samples and a detection model is established by considering the factors of temperature, humidity, carbon monoxide concentration, wind speed and wind direction. Experimental results show that the spiking neural network can achieve a detection accuracy of ∼91%, and therefore provides a better power/accuracy trade-off against existing approaches.