International Journal of Security and Networks (22 papers in press)

Regular Issues

• Safety Assessments of 5G Antenna with Angle Steel Reinforced Transmission Tower based on Full-Scale Test and BP Neural Network  
  by Weizhou Xu, Yashan Hu, Zengju An 
  Abstract: Installing 5G antennas on electrical transmission towers could save cost and space, especially in urban areas. Nevertheless, the safety of the 5G shared transmission tower is rarely reported but is a concerned by engineers. This study performed full-scale experiments on the effects of 5G antennas on the transmission towers, subject to self-weight and wind loads. It revealed that the 5G antennas would not affect the mechanical behaviour of joint areas but the main steel members of transmission towers. Consequently, 5G antennas would probably impair the safety of transmission towers. The effects of 5G antennas on the natural frequency of the transmission towers are investigated by the soil-pile-tower-line-5G finite element model. To facilitate the promotion and quick safety assessment of 5G shared transmission towers, a comprehensive database is constructed based on the finite element model, considering the voltage levels, tangent/tension tower, span lengths, and quantity of 5G antennas. A backpropagation (BP) neural network trained on this database achieves rapid safety evaluations with < 1% relative error for 110 kV towers, enabling engineers to assess 5G-shared towers without complex simulations.
  Keywords: transmission tower; 5G; safety assessment; database; back propagation neural network.
  DOI: 10.1504/IJSN.2025.10071726
   
  
• An Efficient Cooperative Localisation Method Using Sub-network Division and Hybrid Optimisation Algorithm  
  by Luoping Liao, Xiaoyong Yan, Rong Li 
  Abstract: For applications in the internet of things, location information of nodes is important. Conventional localisation methods have three main problems: 1) high communication costs; 2) lower distance measurement accuracy; 3) unstable localisation in uneven node deployments. To this end, this paper proposes an efficient cooperative localisation method using sub-network division and hybrid optimisation. During network initialisation, this method concurrently measures node pair distances using the Salton similarity metric. Using the shared error distribution, we calculate the maximum hop threshold to remove outliers. Next, within the hop count threshold, we divide the entire network into optimal sub-networks. Finally, the hybrid jellyfish and crow search algorithm are introduced to estimate the locations of unknown nodes in sub-networks. Simulation results show that the method significantly outperforms existing approaches in efficiency, accuracy, and stability across various irregular networks. Specifically, the method improves the localisation accuracy by 29.37%93.04% over the current methods.
  Keywords: Cooperative localization; Irregular networks; Similarity metric; Sub-network division; Hybrid optimization algorithm.
  DOI: 10.1504/IJSN.2025.10072004
   
  
• A Network Security Situation Assessment based on Atomic Search Optimised Fuzzy Neural Network  
  by Feiyue Yu 
  Abstract: Modern cyberspace faces threats from sophisticated attacks such as distributed denial-of-service, advanced persistent threats, and zero-day exploits, demanding real-time and adaptive security assessment frameworks. Traditional methods struggle with dynamic threat landscapes due to their limited adaptability to high-dimensional data, susceptibility to fuzzy rule conflicts, and reliance on gradient-based optimization prone to local optima. This study proposes an atomic search-optimised fuzzy neural network model to address the challenges of real-time adaptability, high-dimensional data processing, and fuzzy rule conflicts in network security situation assessment. Integrating quantum-enhanced atom search optimisation with a five-layer fuzzy neural architecture enables the proposed model to dynamically co-optimize fuzzy membership functions and neural network parameters through Lennard-Jones potential-driven population evolution. Key innovations include a fuzzy entropy-based adaptive feature weighting mechanism and a Spark-enhanced atom search optimisation with a five-layer fuzzy neural architecture.
  Keywords: network security situation assessment; Atom Search Optimization; Fuzzy Neural Network.
  DOI: 10.1504/IJSN.2025.10072026
   
  
• A Corrosion Detection Method for Transmission Line Fixtures based on Improved YOLOv8  
  by Chao Liu, Boda Wang, Liying Zhao, Hongxia Ni, Chao Lu, Tonglin Zhang 
  Abstract: In power transmission line maintenance monitoring, hardware corrosion detection is a critical task. However, traditional methods struggle to identify small targets such as metal joints, and the complex environment and limited computing power of the equipment pose additional challenges. To address these issues, this paper proposes a lightweight detection method based on an improved version of You Only Look Once version 8 (YOLOv8), named YOLOv8-Slim-ASFF. This method introduces the GroupShuffleConvolution (GSConv) module to enhance feature extraction capabilities, optimises the neck structure to improve detail capture, and adopts the adaptive spatial feature fusion (ASFF) detection head to enhance localisation accuracy. Experimental results show that its mAP@0.5 is 84.8% higher than that of traditional networks, making it suitable for edge computing devices such as drones, thereby providing support for power transmission line detection.
  Keywords: YOLOv8; rust detection; lightweight network; channel blending; small objective optimization.
  DOI: 10.1504/IJSN.2025.10072466
   
  
• A Survey of Security Issues and Countermeasures for Internet of Medical Things  
  by Tiancheng Chen 
  Abstract: The Internet of Medical Things (IoMT), a network containing connected medical devices with data transferring through the cloud, has become an emerging technology. However, there is an urgent need to enhance security of IoMT. This paper provides abundant knowledge about the IoMT architecture model, typical security issues and countermeasures. The four-layer IoMT architecture model contains layers of perception, network, transport, and application. IoMT security issues includes firmware modification, perception attacks, Sybil attacks, routing attacks, eavesdropping, malware attacks, battery depletion, and Bluetooth attacks. Countermeasures are developed for IoMT systems, such as authentication schemes, anomaly detection, homomorphic encryption, blockchain, federated learning, and intrusion detection systems. Training in artificial intelligence models, blockchain application, quantum computing technology and hardware execution for IoMT security are significant future research directions. Thus, the promising potential of mitigation mechanisms of IoMT security vulnerabilities with embedded frontier techniques is actively altering the era of smart medical empowerment.
  Keywords: internet of medical things; network security; security attacks; artificial intelligence.
  DOI: 10.1504/IJSN.2025.10072499
   
  
• A Tower Tilt Prediction Method for Transmission Lines based on Comprehensive Learning Particle Swarm Optimisation-LSTM and Attention Mechanism  
  by Liying Zhao, Yu Wang, Hongxia Ni, Sinan Shi, Jin Zhu, Tonglin Zhang, Boda Wang 
  Abstract: Transmission tower tilting threatens power grid reliability (e.g., cascading failures, outages, economic losses). Traditional methods (manual inspection, sparse sensors, simple models) suffer high costs, delayed detection, and fail to model complex spatiotemporal dependencies. To address these, a hybrid framework integrating comprehensive learning particle swarm optimisation (CLPSO), LSTM, and attention is proposed, targeting two gaps: deep learning hyperparameter sensitivity and ineffective feature weighting in multivariate time-series. CLPSO globally optimises LSTM hyperparameters to avoid local optima, while attention dynamically weights LSTM output states to focus on critical patterns and enhance multivariate coupling modelling. Experiments show RMSE 0.0008 (60% reduction) and MAE 0.0006 (72.7% reduction) vs. conventional LSTM, with R2 = 0.9629 (explaining 96.29% variance). The framework improves prediction accuracy/robustness for transmission tower intelligent operation and maintenance.
  Keywords: Transmission tower tilt prediction; Comprehensive Learning Particle Swarm Optimization; long short-term memory networks; Attention mechanisms.
  DOI: 10.1504/IJSN.2025.10072539
   
  
• A UAV-enabled, AI-assisted, LiDAR-assisted Fall Prevention System Installation for Transmission Towers  
  by Tao Huang, Zhan Cao, Zijian Song, Chengyu Wang, Chi Wang, Feiyuan Wang, Wei Liu 
  Abstract: Fall protection on energised transmission towers is critical yet challenging. New regulations now mandate 100% attachment of linemen to fall prevention systems during tower ascent. But installation of fall prevention device on in-service towers remains a challenge. To address this, we developed a UAV-assisted system for autonomous installation of temporary anchor devices. A key innovation is a gravity-actuated, self-locking n-shape hook deployed by the UAV to the tower top; it automatically engages to secure a safety line. An ant colony optimisation algorithm computes efficient UAV flight paths between towers, avoiding obstacles with reduced computational load. Compared to manual methods, the system achieved a 73.1% reduction in installation time and 68.8% faster removal. Field trials on live towers completed installations in under 3 minutes, demonstrating enhanced safety and efficiency for high-risk operations. Future work focuses on real-time adaptive path replanning and dynamic obstacle avoidance.
  Keywords: fall protection; transmission towers; UAV; fall prevention device.
  DOI: 10.1504/IJSN.2025.10073086
   
  
• DLARS: A Distributed Localisation Algorithm in Sensor Networks with Automatic Region Segmentation  
  by Rulin Dou, Jiajia Yan, Xiayun Hu 
  Abstract: Most known sensor network localisation approaches assume regular network architectures with nearly linear communication channels. Nonetheless, real-world applications frequently entail irregular and complex node deployments, which result in pathways that deviate significantly from straight lines and cause a notable decrease in placement accuracy. To address this challenge, we propose a distributed localisation algorithm for sensor networks, which divides a irregular network into multiple regular sub-regions. our proposal starts by establishing a reasonable hop count limit based on the error characteristics of anchor nodes. By using this threshold and the distance data between anchor nodes and unknown nodes within this limit, we dynamically create a tailored sub-region for each unknown node. To accurately determine the position of each node, the algorithm employs either a traditional trilateration method or a refined version of the snake optimisation technique, depending on the density of anchor nodes in the vicinity and the spatial relationship between the estimated location and the sub-region boundary. Extensive simulations with various deployment patterns demonstrate that our approach enhances both the accuracy of distance estimation and the reliability of location computation in irregular conditions. Compared to current leading methodologies, our proposed solution improves accuracy by a minimum of 14.59%.
  Keywords: Sensor Network Positioning; t-test; Halton sequence; Snake optimisation.
  DOI: 10.1504/IJSN.2025.10073136
   
  
• Securing Vehicle Network Suspension Control: Lightweight Homomorphic Encryption with Fuzzy Rules  
  by Zefeng Ding, Haili Tang, Xiaojuan Cao 
  Abstract: With the rapid development of the Internet of Vehicles, electromagnetic suspension systems, which ensure ride comfort and handling stability by adjusting stiffness and damping based on sensor data, face critical data tampering threats. Malicious manipulation of data can lead to improper adjustments, excessive vibration, component wear, or even instability. Balancing real-time control, limited onboard resources, and data security is a challenging task. This study proposes an anti-tampering scheme integrating lightweight torus fully homomorphic encryption and fuzzy rule-based control. The torus fully homomorphic encryption method encrypts sensor data in transit, utilising torus fully homomorphic encryption, to ensure confidentiality and integrity while enabling real-time computations. Fuzzy rules handle uncertainty, allowing adaptive decisions on encrypted data. Simulations show that the scheme reduces sprung mass vertical acceleration by 35% compared to unprotected systems, with a total processing latency of 25.8 ms, meeting real-time requirements. It achieves robust tampering resistance, enhanced security.
  Keywords: Internet of vehicles; electromagnetic suspension; data tampering resistance; lightweight homomorphic encryption; fuzzy rules.
  DOI: 10.1504/IJSN.2025.10073590
   
  
• An Insulator Defect Detection Method based on Improved YOLOv8  
  by Yihan Wang, Jingbo Zhang 
  Abstract: This study presents a new method for detecting insulator defects on transmission lines using UAVs in complex environments. Identifying defects like breakage, detachment, and flashover is challenging in practice. The method employs the YOLOv8 framework, incorporating a partial self-attention mechanism in the backbone to capture global context, improving defect detection in cluttered scenes. A convolutional block attention module is added to the neck component to enhance the models representation, boosting detection performance while reducing computational load. Additionally, the standard intersection over union (IoU) loss function is replaced with an enhanced version, improving target localisation and accelerating model convergence. Experiments on a custom insulator dataset showed a 1.4% improvement in mAP@50, as well as increases of 0.4% in precision and 1.1% in recall, compared to the baseline. The results confirm that this method improves the accuracy and robustness of insulator defect detection.
  Keywords: Deep learning; insulator defect detection; self-attention mechanism; target detection; YOLOv8.
  DOI: 10.1504/IJSN.2025.10073610
   
  
• A Topology Network Optimisation of Underground Mine Escape Paths based on Real-Time Personnel Localisation Data  
  by Mingjiang Wu 
  Abstract: Static escape paths often fail to adapt to dynamic underground hazards, including fires and collapses. To address this limitation, we propose a topology-driven optimization framework leveraging real-time personnel tracking. Ultra-wideband positioning data from the subterranean tunnel dataset enable the construction of a spatiotemporal graph network with dynamic risk perception. Our multi-objective cost function integrates path length, personnel density derived from adaptive Gaussian kernel estimation, and hazard gradients simulated via Navier-Stokes equations for fire and predicted through long short-term memory networks for collapses. The dynamic topology optimisation algorithm achieves sub-second path re-planning. Validation using a 500m
  Keywords: underground escape routes; topology network optimisation; personnel location data; dynamic risk perception; mine safety.
  DOI: 10.1504/IJSN.2025.10073839
   
  
• Multi-Agent Deep Reinforcement Learning-Driven Nash Equilibrium Computation for Supply Chain Networks  
  by Wenhui Li, Jiawen Zhang, Can Wang, Ya Li 
  Abstract: The stable computation of Nash equilibrium in supply chain networks is fundamental to achieving optimal multi-agent coordination and system efficiency. Focusing on issues of poor adaptability and slow convergence in existing methods, this paper first develops a hybrid reward mechanism that coordinates individual and system-level objectives through a game-theoretical formulation. Secondly, the equilibrium computation process is optimised using an adaptive policy synchronisation protocol, ensuring stable convergence under dynamic market conditions. Finally, the multi-agent deep reinforcement learning framework is implemented with performance evaluation metrics including inventory turnover and cost efficiency. Experimental validation demonstrates consistent performance improvements: convergence in just 21 iterations versus the conventional 100, 0.94 inventory turnover, and robust 0.86 order fulfilment with 0.78 cost efficiency, all surpassing the 0.70 baseline. The proposed solution offers immediate value for real-world supply chain optimisation, particularly in manufacturing and logistics, where dynamic coordination is crucial.
  Keywords: multi-agent deep reinforcement learning; supply chain network optimisation; Nash equilibrium solution; dynamic market environment.
  DOI: 10.1504/IJSN.2025.10073850
   
  
• Submarine Cable Monitoring based on Regularised Machine Learning and Fibre Optic Vibration Sensors  
  by Qizhi Bian, Boqing Li, Bingxi Chen, Jiyin Shi 
  Abstract: Submarine cables are critical for global energy transmission and intercontinental communication. Traditional monitoring methods include manual patrols, which are costly and limited by weather and sea conditions, failing to achieve continuous monitoring. Additionally, acoustic detection is easily interfered with by underwater noise, leading to low accuracy in identifying micro-disturbances. These issues result in low real-time performance and high false alarm rates. To address this, this paper proposes a system that combines fibre optic vibration sensors and regularised machine learning. Fibre optic sensors enable long-distance, distributed monitoring with high sensitivity and anti-interference capabilities, while regularised learning enhances stability. The system uses a four-layer architecture, extracts time-frequency features, and applies L1 regularisation shrinking redundant feature weights to zero for selection with a particle swarm optimisation regularised support vector machine. Experimental results demonstrate 94.2% accuracy, a 2.8% false alarm rate, and a delay of less than 0.3 seconds.
  Keywords: submarine cable monitoring; fibre optic vibration sensor; regularised machine learning; signal processing; early warning system.
  DOI: 10.1504/IJSN.2025.10073860
   
  
• Guarding the Gap: A Machine Learning Framework for Covert Storage Channel Detection  
  by Rajesh Gaikwad, Dr. Dhananjay Dakhane, Pratik Jalan 
  Abstract: The primary issue of covert cyber defense channels is a new concern, as they allow illicit data to go undetected with traditional security measures. In response, a new machine learning solution is proposed, combining domain-aware feature extraction and ensemble learning to enhance the detection of hidden storage channels. The solution identifies indicators like file access patterns, metadata anomalies, and entropy variation, allowing accurate classification into legitimate and hidden behavior. A three-pronged hybrid ensemble of Decision Trees, SVM, and DNN was employed, with detection rates exceeding 99.3% on a large-scale simulated dataset. Adversarial training has also been applied to enhance model generalizability, making it more evasion-resistant to advanced attacks like metadata hiding and access pattern manipulation. The framework is computation-efficient and adjustable in real-time for SIEM systems. Future work should focus on real-world data verification and unsupervised learning methods for discovering new covert communication patterns.
  Keywords: covert storage channel; anomaly detection; machine learning; metadata analysis; entropy analysis; adversarial learning; cybersecurity.
  DOI: 10.1504/IJSN.2025.10074728
   
  
• Static Analysis of Security Vulnerabilities in Android-Based Academic Information Systems for Higher Education in Indonesia  
  by Nashrul Hakiem, Hadid Syaifullah Albab, Nenny Anggraini, Sandra H. Afrizal, Imam Marzuki Shofi, Luh Kesuma Wardhani 
  Abstract: Smartphones mark a significant technological revolution, but recent years have seen a surge in data leaks. Academic information systems play a crucial role in managing academic information on campuses. Indonesia witnesses a growing adoption of application-based academic information systems due to their accessibility, emphasising the need to secure these systems. This study aims to develop a WhatsApp chatbot for security testing of Android applications using static analysis. The chatbot, built on the WhatsApp platform using Node.js and application programming interfaces from the mobile security framework and VirusTotal, reveals that among 18 applications, 15 have a medium security risk, two exhibit low risks, and one presents a high risk. Common vulnerabilities include weak cryptography, dangerous permissions, and hardcoded secrets. Furthermore, several vulnerabilities align with open web application security project top ten mobile vulnerabilities, encompassing insecure data storage, insufficient cryptography, client code quality, and reverse engineering.
  Keywords: Android applications; Mobile Security framework; Academic Information System; WhatsApp chatbot; Static analysis.
  DOI: 10.1504/IJSN.2025.10074839
   
  
• Dynamic Multi-Factor Authentication Algorithm for Secure Microservice Architecture in Digital Capability Open Platforms  
  by Shujun Nan, Xinyi Liu, Shan Wei, Zhizhi Zhang 
  Abstract: To address the security challenges of identity authentication in microservice-based digital open capability platforms, this paper proposes a dynamic multi-factor authentication algorithm. The algorithm adaptively selects and combines authentication factors based on real-time risk levels, enabling frictionless, lightweight authentication for low-risk access while enforcing strong multi-factor verification in high-risk scenarios. A stateless policy execution engine is implemented to support real-time authentication decisions and the issuance of tokens. Experimental results demonstrate that the algorithm effectively intercepts 98.1% of credential theft attacks, ensuring the security of core platform resources. It also maintains a 98.7% authentication success rate under high concurrency of 20,000 users, significantly improving access experience for legitimate users while providing a flexible and adaptive security mechanism for open platforms.
  Keywords: digital capability open platform; microservices architecture; dynamic multi-factor authentication; risk awareness.
  DOI: 10.1504/IJSN.2025.10075122
   
  
• Machine Learning Based Efficient Prevention Technique for XSS Cyber Security Attack  
  by Barasha Das, Liaren Emani Aier, Sreya Bhowmick, Arpita Nath Boruah, Mrinal Goswami 
  Abstract: Cybersecurity is an essential barrier to defending digital systems and data against many cyber threats, including the well-known Cross-Site Scripting (XSS) attack. XSS occurs when malicious scripts are inserted into web pages for unauthorized actions or data theft. This paper presents a comprehensive approach to combating XSS attacks by integrating detection and prevention models leveraging machine learning techniques. Feature selection is conducted using an autoencoder, enhancing model efficiency. The detection model employs a stacking methodology, while the prevention method integrates cybersecurity tools such as WAF, IDS, IPS, and RASP. Output from the stacking model and pre-processed data serve as input for the prevention model, ensuring a robust defence mechanism against XSS attacks. The proposed XSS prevention and detection efforts aim to extract features effectively from the dataset, enhancing the model's overall performance. The proposed model demonstrates a rapid processing speed and high evaluation matrix results, indicating its efficiency and effectiveness.
  Keywords: Cross-Site Scripting (XSS); Cybersecurity; Detection; Machine Learning; Prevention.
  DOI: 10.1504/IJSN.2025.10075171
   
  
• Multiscale Fusion Dehazing Algorithm Based on Channel Attention Mechanism with Sub-pixel Convolution  
  by Chao Lu, Hongxia Ni, Hang Jiang, Qi Luan, Entie Qi, Xue Li 
  Abstract: This study aimed to propose a multiscale fusion dehazing algorithm based on the channel attention mechanism with sub-pixel convolution (MCASDA) to address the issues of blurred image details, colour distortion, and artefact residue in complex haze scenes. The algorithm was based on the convolutional networks for biomedical image segmentation architecture and achieved multilevel feature co-optimisation through encoder and decoder frameworks. The enhanced residual network was designed with the convolutional layer of the Swish activation function in the encoder stage. Combined with the channel attention mechanism, thus significantly improving the ability of the network to focus on key features. Sub-pixel convolutional upsampling was used to replace the traditional transpose convolution in the decoder stage, avoiding checkerboard artefacts through pixel rearrangement. Moreover, the multiscale feature fusion module was combined with back-projection feedback to recover high-frequency details and spatial information effectively. The evaluation outcomes indicated that MCASDA achieved notable improvements in peak signal-to-noise ratio and structural similarity index on the synthetic objective testing set dataset (29.09 dB and 0.9632, respectively), surpassing conventional approaches. These findings substantiated the algorithm's superior performance in dehazing, thereby effectively producing higher-quality dehazing images with enhanced clarity.
  Keywords: channel attention mechanism; feature fusion; image dehazing; sub-pixel convolution; biomedical image segmentation network architecture.
  DOI: 10.1504/IJSN.2025.10076359
   
  
• Phishing Websites Detection based on Metaheuristic Optimisation and Ensemble Algorithms  
  by Kayode Sakariyah Adewole, Moshood A. Hambali, Yakub Saheed, Victor E. Adeyemo, Kamaldeen J. Muhammed, Rafiu M. Isiaka 
  Abstract: Phishing attacks pose a significant threat to web- and mobile-based applications, aiming to compromise network security and hijack legitimate resources. Illegitimate websites, fake emails, and deceptive web/mobile solutions are used to collect user credentials and compromise accounts. Traditional blacklist-based methods for detecting phishing websites suffer from a lack of zero-day attack detection. To address this, machine learning (ML) models have been widely explored. However, meta-heuristic approach that covers studies on feature selection to create compact ML models for phishing detection has not been extensively studied. This paper fills that gap by analysing ten meta-heuristic algorithms to identify discriminating features for detecting phishing websites. The study evaluates three ensemble-based classifiers - Random Forest, AdaBoost, and Rotation Forest. The results show that Firefly Optimization Algorithm (FOA) produced the best results Using the features selected by FOA, Random Forest outperforms other ML algorithms with 95.36% accuracy and AUC-ROC of 98.9%. This result is followed by rotation forest (accuracy 94.89%) and AdaBoost performs the least with accuracy of 91.42%. The findings highlight the efficacy of the FOA meta-heuristic algorithm in phishing website detection.
  Keywords: Phishing; security; meta-heuristic; machine learning; ensemble algorithms.
  DOI: 10.1504/IJSN.2025.10076529
   
  
• Optimization of Underwater Inspection for Offshore Wind Power Facilities Using Multimodal Data Fusion and Deep Networks  
  by Hao Wu, Jiakun Wang 
  Abstract: Underwater inspection is vital for the safety of offshore wind power facilities, yet it faces significant challenges. Traditional methods using single sensor types are unreliable: optical images degrade in murky water, while acoustic data lacks fine detail. This research overcomes these limitations by intelligently fusing both optical and acoustic sensor data. A dynamic weighting mechanism adaptively combines their strengths at the feature level, ensuring robust performance in changing water conditions. Furthermore, we introduce a meta-reinforcement learning planner to create a closed-loop perception-decision system. This enables autonomous vehicles to adjust their inspection path in real-time based on live sensor feedback, focusing efficiently on potential defect areas. Extensive evaluation demonstrates that our method achieves a mean accuracy of 96.5% in defect detection and reaches an area coverage of 98% in only 382 steps, significantly outperforming conventional approaches. This offers a practical and efficient solution for inspecting intelligent underwater infrastructure.
  Keywords: multimodal fusion; offshore wind power; underwater inspection; meta-reinforcement learning; dynamic weight assignment.
  DOI: 10.1504/IJSN.2025.10076825
   
  
• Identity-based Dynamic Provable Data Possession with Public Verification for Multi-replica Cloud Storage  
  by Linmei Jiang, Zhengtian Lu 
  Abstract: To address the growing demand for highly reliable multi-replica storage in cloud computing and the lack of effective dynamic multi-replica Provable Data Possession (PDP) schemes, this paper proposes an identity-based PDP method that supports dynamic data operations. The scheme allows users to insert, delete, and modify data without needing to download and re-upload the entire file, thereby significantly improving data processing efficiency. In addition, the proposed method supports public verification, enabling users to delegate data integrity checks to a semi-trusted third party without compromising privacy. Technically, the scheme enhances performance by incorporating skip lists to improve computational efficiency, random masking to encrypt multiple replicas, and homomorphic tags to aggregate verification data, thereby reducing communication overhead. Experimental results demonstrate that the proposed method achieves high computational efficiency, low communication cost, and minimal storage overhead.
  Keywords: Provable data possession; Multiple replicas; Public verification; Data security; Cloud storage; Dynamic data update.
  DOI: 10.1504/IJSN.2025.10076885
   
  
• Supply Chain Demand Forecasting based on Gated Graph Neural Networks and Federated Learning  
  by Wenhui Li, Shang Xue, Can Wang, Ya Li 
  Abstract: Facing the challenge of data silos among enterprises in supply chains while requiring collaborative forecasting, this study proposes an innovative approach integrating gated graph neural networks with federated learning. This combination allows the model to effectively capture complex supply chain relationships while maintaining data privacy through decentralised training. By leveraging graph structure to analyse inter-enterprise networks and employing federated learning to enable joint model training without sharing raw data, our method successfully addresses both prediction accuracy and privacy concerns. Extensive validation on multiple real-world supply chain datasets demonstrates that this approach outperforms conventional methods, improving prediction accuracy by 12.5% and reducing error by 15.3%, thereby providing a practical solution for building secure and efficient collaborative forecasting systems in supply chain management.
  Keywords: supply chain demand forecasting; neural network for gate control charts; federated learning; data privacy protection.
  DOI: 10.1504/IJSN.2025.10077155