In this paper, we propose an enhanced clustering protocol that integrates an improved K-means with a Mobility-Aware Cluster Head-Election Scored (IK-MACHES) algorithm, designed for extending the lifetime and operational efficiency of Wireless Sensor Network (WSN) with mobility. Variety approaches applying Low Energy Adaptive Clustering Hierarchy (LEACH) often struggle to manage optimal energy distribution due to their static clustering and limited cluster head (CH) selection criteria, primarily focusing on the proximity of residual energy or distance. Thus, this paper proposes an algorithm that takes into account both the residual energy of sensor nodes and the distance between the cluster’s central point to the base station (BS), which ultimately enhances the network’s lifetime. Additionally, our approach incorporates mobility considerations, enhancing the adaptability of the mobility environments, such as autonomous vehicular networks. Our proposed method first constructs the cluster’s configuration and then elects the CH applying an improved K-means clustering algorithm—one of the machine learning methods—integrated with a proposed IK-MACHES mechanism. Three CH scoring strategies in the proposed IK-MACHES protocol evaluate the residual energy of the nodes, their distance to the BS and the cluster central point, and relative node’s mobility. The simulation results demonstrate that the proposed approach improves performance in terms of the first node dead (FND) and 80% alive nodes metrics with mobility, compared to other LEACH protocols such as classical LEACH, LEACH-B, Improved-LEACH, LEACH with K-means, Particle Swarm Optimization (PSO), and LEACH-GK protocol, thereby enhancing network lifetime through optimal CH selection.

Traffic accident investigation has been used to reconstruct the cause and condition of the accident by using images and driving record data that introduce IT technology, and to analyze both sides of the accident, skid marks, and vehicle damage to identify the perpetrators and victims. However, level 3 self-driving vehicles are the most important factor in determining the cause and imputation of the accident by the driver or manufacturer with control information at the time of the accident. It is also developing into a network and connected vehicle with various restrictions such as the burden of the price of sensors equipped with self-driving cars and climate and rapidly changing road traffic information. In addition, network and connected vehicle driving data are stored on the outside, or various devices and sensors are installed to store information on the outside for convenience in operation, and efforts to enact laws are continuing. This paper attempts to propose a traffic investigation digital framework using digital data generated by these devices and sensors.

Cyber-physical system (CPS) incorporates several computing resources, networking units, interconnected physical processes, and monitoring the development and application of the computing system. Interconnection between the cyber and physical worlds initiates attacks on security problems, particularly with the enhancing complications of transmission networks. Despite the efforts to combat these problems, analyzing and detecting cyber-physical attacks from the complex CPS is challenging. Machine learning (ML)-researcher workers implemented based techniques to examine cyber-physical security systems. A competent network intrusion detection system (IDS) is essential to avoid these attacks. Generally, IDS uses ML techniques to classify attacks. However, the features used for classification are not frequently appropriate or adequate. Moreover, the number of intrusions is much lower than that of non-intrusions. This research presents an African Buffalo Optimizer Algorithm with a Deep Learning Intrusion Detection (ABOADL-IDS) model in a CPS environment. The main intention of the ABOADL-IDS model is to utilize the FS with an optimal DL approach for the intrusion recognition and identification procedure. Initially, the ABOADL-IDS model performs the data normalization process. Furthermore, the ABOADL-IDS model utilizes the ABO technique for feature selection. Moreover, the stacked deep belief network (SDBN) technique is employed for intrusion detection and identification. To improve the SDBN technique solution, the seagull optimization (SGO) technique is implemented for the hyperparameter selection. The assessment of the ABOADL-IDS technique is accomplished under NSLKDD2015 and CICIDS2015 datasets. The performance validation of the ABOADL-IDS technique illustrated a superior accuracy value of 99.28% over existing models concerning various measures.

At present, cyber-attacks have become more critical and familiar, which appeals to a novel line of security defences to defend against them. Cyber Threat Intelligence (CTI) originated as a reputation in the frequently developing cybersecurity landscape, vital in safeguarding digital models. Understanding this domain develops over a complete study of intelligence’s numerous features and sources. Data sharing and analysis centres perform as alarms of collaboration, demonstrating the collective vigilance needed to oppose such developing attacks. Inspecting and gathering data about cyberattacks from tweets can effectively deliver critical perceptions of the threats, their effects, occurrence areas, and probable mitigation tactics. Existing study on cyberattack absences in establishing Artificial Intelligence (AI) based analytic solutions for delivering country-wide cyber-attack intelligence. Cyber planners at a domestic level need AI-based decision support models to determine a country’s cyber attitude or vigilance. This study designs and develops an Enhanced Threat Intelligence for Cybersecurity Using an Ensemble of Deep Leaning Models with Metaheuristic Optimization Algorithm (ETIC-EDLMOA) model. The presented ETIC-EDLMOA model’s main aim is to detect and mitigate network attacks in cybersecurity effectively. Initially, the ETIC-EDLMOA model undergoes a data pre-processing stage to ensure clean and structured input data for analysis. Besides, the Word2vec model is utilized for feature extraction. For the classification process, the ensemble of DL models is employed, including the recurrent neural network (RNN) method, long short-term memory (LSTM) model, and conditional variational autoencoders (CVAE) technique. Finally, the ensemble models’ hyperparameter fine-tuning process is performed using the Wolverine optimization algorithm (WoOA) technique. A comprehensive range of simulation analyses is conducted to ensure the improved performance of the ETIC-EDLMOA method on the CybAttT dataset. The comparison study of the ETIC-EDLMOA method illustrated a superior accuracy value of 98.51% over existing techniques.

The gallbladder is a small, hollow organ positioned beneath the liver, primarily responsible for temporarily storing bile. Bile is a fluid formed by the liver that helps digestion. There are various types of gallbladder disease. Early diagnosis and identification are crucial for effective treatment of gallbladder disorders. Poor medical outcomes and improved patient symptoms may result from errors or delays in diagnosis. Numerous symptoms and signs, particularly those associated with gallbladder disorder, might be blurred. As a result, medical specialists must understand and interpret ultrasound images. Considering that ultrasound imaging for diagnosis is labour- and time-consuming, it might be difficult to support financially. Deep learning (DL) is an effective model that could help early identify gallbladder disease using ultrasound (US) images. In this paper, a Hybrid Deep Learning Model with Feature Engineering for the Accurate Diagnosis of Gallbladder Disease Types (HDLMFE-ADGDT) approach is proposed. The primary purpose of the HDLMFE-ADGDT approach is to develop an effective DL-based method for classifying GB disease categories. As an initial step, the HDLMFE-ADGDT technique employs a Non-Local Means (NLM) filter to remove noise and enhance the overall image quality. For feature extraction, the Squeeze-and-Excitation Capsule Network (SE-CapsNet) is employed. At last, the hybrid convolutional neural network with bidirectional long short-term memory (CNN-BiLSTM) is implemented for gallbladder disease diagnosis. A comprehensive set of experiments is conducted to validate the performance of the HDLMFE-ADGDT model on the Gallbladder diseases dataset. The empirical results demonstrate that the HDLMFE-ADGDT model outperformed existing methodologies, achieving 99.09% accuracy, 95.83% precision, 95.87% sensitivity, and 99.49% specificity.

Harris Hawk Optimizer (HHO) is a recent revolutionary algorithm developed in the literature that simulates the cooperative hunting behaviour of Parabuteo Unicinctus. Despite its simplicity, the standard HHO often suffers from slow convergence, limited exploitation capacity and performance degradation on high-dimensional and constrained problems. This study aims to develop seven novel Harris Hawk Optimizer (HHO) variants, HHO-ADAP, HHO-CHAOS, HHO-Elite, HHO-GA, HHO-Inertia, HHO-PSO, and HHO-ULTRA, that integrate adaptive mechanisms, chaotic dynamics, elite preservation, and cross-algorithmic hybridization to improve the balance between exploration and exploitation. The proposed methods were rigorously tested on the CEC 2014 benchmark suite for dimensions 10, 30, 50, and 100, as well as ten constrained engineering design problems, and results are compared against state-of-the-art optimizers CMA-ES, L-SHADE, LSHADE-cnEpSin, SPS-L-SHADE-EIG, EBOwithCMAR, WMA, and OWMA. Quantitative results demonstrate that the hybrids consistently outperform the baseline HHO and classical optimizers. HHO-PSO and HHO-Elite achieved up to 35% faster convergence and reached solution values as small as 10⁻²¹⁶, compared with much weaker values (10⁻⁴²–10⁻⁴⁷) for classical baselines. On multimodal and fixed-dimension functions, HHO-Elite, HHO-CHAOS, and HHO-ADAP effectively delayed stagnation and preserved diversity, avoiding premature convergence. For engineering problems, the hybrids produced near-optimal designs: pressure vessel (≈5885.2), spring (≈0.01267), welded beam (≈1.7257), gear train (= 0), and Belleville spring (≈1.9795). Variance was as low as 10⁻¹⁶ (multiple disk clutch, gear train), while average runtimes remained below 0.01 s for most hybrids, markedly faster than champion algorithms such as SPS-L-SHADE-EIG (> 1.4 s) and WMA (> 1.8 s). The results highlight that hybridization significantly enhances HHO’s robustness, solution accuracy, and adaptability for solving large-scale, nonlinear, and constrained optimization problems in engineering and scientific domains.