In today’s digital landscape, email is acknowledged as a critical conduit for global data exchanges. With a surge in data volume, malefactors exploit user identities, leading to data misuse. Cybercriminals employ electronic transgressions such as phishing and spam to orchestrate security infractions. Machine learning counters these breaches using myriad techniques, demonstrating significant efficiency in identifying phishing emails. We can divide machine learning into two types: supervised and unsupervised. Supervised learning requires pre-training the model on labelled datasets, amalgamating classification, and regression learning. Notably, supervised methodologies such as support vector machines (SVMs), naive Bayes, decision trees, neural networks, random forests, and deep learning have been exploited for spam filtering. This review delves into issues concerning spam filtering and email classification through supervised machine learning techniques, offering a comprehensive evaluation of strategies, methods, performance indicators, and the benefits and drawbacks of different research. This information allows researchers to assess the efficiency and effectiveness of supervised learning algorithms, laying the foundation for advanced email categorization techniques.

The misuse of privileges by users can lead to significant reputational and financial losses for enterprises. To reduce the risk of information leakage, it is crucial to detect and analyze abnormal behaviours of internal employees. Firstly, based on the characteristics of internal employee behaviour, a data filter strategy based on user behaviour is proposed. Then, a data balancing strategy based on the concept of hybrid sampling is introduced. Moreover, to further construct the behaviour model, an improved particle swarm optimization algorithm based on adaptive delay and genetic factors is proposed, and it is used to search for the optimal parameters of LightGBM. Experimental results demonstrate that the proposed method is highly effective in detecting internal threats.

Deep reinforcement learning (RL) approaches are increasingly prominent in the field of robotics due to their adaptive decision-making capability. However, developing a single RL agent capable of performing multiple continuous control tasks for quadruped robots remains challenging. In this paper, a multi-task deep RL framework based on curriculum learning and policy distillation is proposed, which aims to enhance the quadruped robot's motor performance across multiple continuous tasks by leveraging knowledge from expert skill teachers. The main novelties of the framework lie in the self-optimizing terrain curriculum learning strategy and the improved distillation loss function. The proposed self-optimizing terrain curriculum learning strategy for quadrupedal robots is designed to utilize Bayesian optimization to predict potential training terrains, thus effectively identifying the most suitable training curriculum. Additionally, the improved distillation loss function for RL weight optimization is proposed to enhance the transferability of the trained policy across diverse tasks. To validate the effectiveness of the proposed multi-task deep RL framework, the performance of the policy generated by the framework across diverse terrains is assessed. The experimental results demonstrate that the proposed multi-task deep RL framework could generate a unified policy that achieves excellent performance across multiple continuous control tasks for quadruped robots.

This study addresses the problem of centralized fusion filtering for β-quaternion signals obtained from multisensor observations. It is assumed that observations from each sensor may be subject to random updates or delays, and may also experience packet losses. In such cases, a compensation strategy is employed, where the missing observation is replaced by its predictor. Furthermore, the correlation between the additive noise of the signal and the observations is taken into account. Under first-order properness conditions, a recursive algorithm is developed to compute the optimal filter. A key feature of the proposed algorithm is its ability to reduce the dimensionality of operations in each iteration by a quarter of that required by the optimal β-quaternion processing algorithm, which relies on augmented processes. This reduction significantly enhances computational efficiency while maintaining identical estimators under first-order properness conditions. A numerical example is presented to illustrate, among other aspects, the substantial reduction in computational time achieved by the proposed algorithm, as well as its superior estimation accuracy compared to conventional quaternion processing methods.

The complex cross-region transfer relationship of carbon emissions from transportation increases the difficulty of achieving the overall goal of carbon emission reduction. This paper calculates the transportation carbon emissions from 2011 to 2020 in the Pearl River Delta. Based on an improved gravity model, a spatiotemporal relationship network for cross-region transfer is constructed. The overall characteristics, individual characteristics, and block models of the network are analysed with the social network analysis method. Through these efforts, the spatiotemporal relationship of cross-region transfer of carbon emissions is depicted. The findings indicate that the Pearl River Delta exhibits complex spatiotemporal patterns of transportation-related carbon emissions. With increasing network density and stable spatial correlations, cross-regional transfers are facilitated. Core cities like Guangzhou and Shenzhen play pivotal roles, while distinct city blocks exhibit stable yet varied participation. This highlights the need for coordinated, cross-regional emission reduction mechanisms that consider geographical and economic factors.

With the rapid advancement of information technology, the networked systems (NSs) have found extensive application across various fields. As the complexity and scale of NSs continue to increase, the fault detection plays an important role in ensuring the reliable operation of the system and has attracted the attention of scholars. In this paper, the research progress of fault detection (FD) for NSs is reviewed, especially the FD methods in communication-constrained situations. First, the research significance of FD for NSs and the current research hotspots are reviewed. Subsequently, the methods of FD subject to time delay, measurement loss, signal quantization and cyber attack are summarized. Next, for the NSs constrained by communication protocols, the corresponding FD methods are recalled. Then some research results on FD for NSs with finite-time index constraint and memory-scheduled strategy are presented. Finally, the limitations of current FD methods are summarized and the future research directions are pointed out.

This research addresses challenges in capsicum peduncle detection in night-time greenhouse environments, including low light, uneven illumination, and shadows, using advanced computer vision models. A dataset of 200 images was curated, capturing diverse distances, heights, occlusion levels, and lighting conditions, and was rigorously pre-processed and augmented. Two YOLOv9 instance segmentation variants, YOLOv9c-seg and YOLOv9e-seg, were custom-trained and fine-tuned using Google Colaboratory. YOLOv9c-seg (56.3 MB) achieved superior mean Average Precision (mAP) scores of 0.751 (box) and 0.725 (mask), outperforming YOLOv9e-seg (121.9 MB) with mAP scores of 0.674 (box) and 0.658 (mask). Grounded SAM, a zero-shot segmentation model, achieved maximum peduncle detection confidences of 59% and 49% with positional prompts. Comparative testing on 50 images containing 70 capsicums showed YOLOv9c-seg achieving mean precision, recall, and F1-scores of 0.93, 0.86, and 0.89, respectively, outperforming Grounded SAM (0.86, 0.70, and 0.77). This study highlights the efficacy of single-shot versus zero-shot segmentation models for automated capsicum peduncle detection in controlled agricultural environments, offering insights into model performance and future research directions for model optimization and dataset expansion.

Diabetic retinopathy (DR) is a major cause of blindness among adults worldwide. Detecting and classifying DR early is essential for timely treatment and prevention of vision loss. This study introduces a new approach to identify and classify DR by using federated learning (FL) environment and Federated differential evolution optimization (FedDEO). FL enables collaborative learning across multiple decentralized devices while maintaining data privacy. FedDEO optimization enhances the model's performance by fine-tuning hyperparameters in a distributed manner. The research achieves state-of-the-art accuracy in DR classification using the MESSIDOR dataset for training and testing. The findings suggest that integrating FedDEO optimization within the FL framework can significantly improve DR detection, offering a scalable and privacy-preserving solution for ophthalmic healthcare. The FedDEO algorithm optimizes hyperparameters, including learning rates and batch sizes, to enhance model performance. The experiments show that FedDEO works effectively for the MESSIDOR dataset to achieve a high classification accuracy, specificity, recall, and F1-Measure of 96.98%, 98.12%, 97.12%, and 98.00% while preserving privacy. The results demonstrate that implementing these algorithms in an FL environment significantly improves privacy and performance compared to other approaches. Specifically, the proposed technique outperforms others in terms of accuracy, specificity, recall, and F1 score when applied to categorizing retinal images.

The extended line of sight (LOS) guidance law is designed to track toward a target in a three-dimensional space, and this method presents a family of guidance laws resulting in a rich behaviour for different parameters. First, polar kinematics models for the pursuer and target are established by taking the observer as a reference point. Second, the extended LOS guidance law was deduced through polar kinematics models. Third, information fusion weighted by matrices was applied to target tracking, providing a good estimate for extended LOS guidance. This combination can enrich the application range of guidance laws and information fusion. Finally, the simulations were implemented in MATLAB. The simulation results confirm the availability of this guidance law.

Modular Multilevel Converters (MMCs) play a crucial role in high-voltage DC (HVDC) systems due to their adaptable control and swift response capabilities. However, AC side faults could introduce asymmetric double-line frequency components into the MMC's circulating current, jeopardizing DC bus stability. Although strategies to mitigate DC side fluctuations are well-established, a detailed comparison of their operational impacts remains unexplored. The operational effects of MMCs under AC side faults are firstly examined in this study, including issues related to grid-connected current asymmetry and DC bus current fluctuations. Subsequently, two targeted control schemes are devised based on the goals of suppressing the double-line frequency components and only the zero-sequence components of circulating current. Through simulation, these approaches are contrasted, delineating their advantages and disadvantages in terms of circulating current suppression, submodules(SMs) capacitance voltage fluctuation, and other performance metrics, which could provide a foundational reference for designing and selecting MMC control strategies under AC side faults.