To ensure no private information is leaked in the aggregation phase in federated learning (FL), many frameworks use homomorphic encryption (HE) to mask local model updates. However, the heavy overheads of these frameworks make them unsuitable for cross-device FL, where the clients are a huge number of mobile and edge devices with limited computing resources. Even worse, some of them also fail to manage the dynamic changes of clients. To overcome these shortcomings, we propose a threshold multi-key HE scheme tMK-CKKS and design an efficient and robust privacy-preserving FL framework. Robustness means that our framework allows clients to join in or drop out during the training process. Besides, because our tMK-CKKS scheme can pack multiple messages in a single ciphertext, our framework significantly reduces the computation and communication overhead. Moreover, the threshold mechanism in tMK-CKKS ensures that our framework can resist collusion attacks between the server and no more than t (threshold value) curious internal clients. Finally, we implement our framework in FedML and conduct extensive experiments to evaluate our framework. Utility evaluations on 6 benchmark datasets show that our framework can protect privacy without sacrificing the model accuracy. Efficiency evaluations on 4 typical deep learning models demonstrate that: our framework can speed up the computation by at least 1.21 $$\times $$× over xMK-CKKS-based framework, 15.84 $$\times $$× over Batchcrypt-based framework, and 20.30 $$\times $$× over CRT-Paillier-based framework. Our framework can reduce the communication burden by at least 8.61 MB over Batchcrypt-based framework, 35.36 MB over xMK-CKKS-based framework and 42.58 MB over CRT-Paillier-based framework. The advantages in both computation and communication expand with the size of deep learning models.

Pseudorandom number generators are deterministic algorithms capable of producing sequences of numbers that appear sufficiently ”random,” and they find extensive applications across various domains such as cryptography, network security, communications, machine learning, and gaming. As highly nonlinear mathematical systems, neural networks exhibit characteristics such as fitting ability, unidirectional property, generalization capability, and parallelism, which render them prominent in the design of PRNGs and a topic of significant research interest. To date, there exists no comprehensive review focusing on the utilization of neural networks for the design of PRNGs. This paper categorizes existing neural network-based PRNG design schemes into three types: those based on recurrent neural network models and their variants, such as Long Short-Term Memory (LSTM) models; those based on generative adversarial networks (GANs); and those based on deep reinforcement learning. Subsequently, the paper elucidates the design philosophies and typical algorithmic principles underlying these schemes, compares these algorithms, summarizes the existing challenges, and discusses prospective research directions.

This article focuses on the distributed cooperative compound tracking issue of the vehicular platoon. First, a definition, called compound tracking control, is proposed, which means that the practical finite-time stability and asymptotical convergence can be simultaneously satisfied. Then, a modified performance function, named finite-time performance function, is designed, which possesses the faster convergence rate compared to the existing ones. Moreover, the adaptive neural network (NN), prescribed performance technique, and backstepping method are utilized to design a distributed cooperative regulation protocol. It is worth noting that the convergence time of the proposed algorithm does not depend on the initial values and design parameters. Finally, simulation experiments are given to further verify the effectiveness of the presented theoretical findings.

Pseudo-random number generators (PRNGs) are deterministic algorithms that generate sequences of numbers approximating the properties of random numbers, which are widely utilized in various fields. In this paper, we present a Genetic Algorithm Optimized Generative Adversarial Network (hereinafter referred to as GAGAN), which is designed for the effective training of discrete generative adversarial networks. In situations where non-differentiable activation functions, such as the modulo operation, are employed and traditional gradient-based backpropagation methods are inapplicable, genetic algorithms are utilized to optimize the parameters of the generator network. Based on this framework, we propose a novel recursive PRNG. Given that a PRNG can be constructed from one-way functions and their associated hardcore predicates, our proposed generator consists of two neural networks that simulate these functions and serve as the state transition function and the output function, respectively. The proposed PRNG has been rigorously tested using stringent benchmarks such as the NIST Statistical Test Suite (SP800-22) and the Chinese standard for random number generation (GM/T 0005-2021). Additionally, it has demonstrated outstanding performance in terms of Hamming distance. The results indicate that the proposed GAN-based PRNG has achieved a high degree of randomness and is highly sensitive to variations in the input.

In recent years, social network sentiment classification has been extensively researched and applied in various fields, such as opinion monitoring, market analysis, and commodity feedback. The ensemble approach has achieved remarkable results in sentiment classification tasks due to its superior performance. The primary reason behind the success of ensemble methods is the enhanced diversity of the base classifiers. The boosting method employs a sequential ensemble structure to construct diverse data while also utilizing erroneous data by assigning higher weights to misclassified samples in the next training round. However, this method tends to use a sequential ensemble structure, resulting in a long computation time. Conversely, the voting method employs a concurrent ensemble structure to reduce computation time but neglects the utilization of erroneous data. To address this issue, this study combines the advantages of voting and boosting methods and proposes a new two-stage voting boosting (2SVB) concurrent ensemble learning method for social network sentiment classification. This novel method not only establishes a concurrent ensemble framework to decrease computation time but also optimizes the utilization of erroneous data and enhances ensemble performance. To optimize the utilization of erroneous data, a two-stage training approach is implemented. Stage-1 training is performed on the datasets by employing a 3-fold cross-segmentation approach. Stage-2 training is carried out on datasets that have been augmented with the erroneous data predicted by stage 1. To augment the diversity of base classifiers, the training stage employs five pre-trained deep learning (PDL) models with heterogeneous pre-training frameworks as base classifiers. To reduce the computation time, a two-stage concurrent ensemble framework was established. The experimental results demonstrate that the proposed method achieves an F1 score of 0.8942 on the coronavirus tweet sentiment dataset, surpassing other comparable ensemble methods.