ABSTRACT With the rapid development of connected autonomous vehicles, intelligent traffic systems are increasingly integrating Roadside Unit (RSU). However, some edge devices in the RSU telecommunication network with limited computing power, due to their simple structures, struggle to implement complex security measures, making them vulnerable nodes in RSU networks and prime targets for cyberattacks. Hence, technologies such as intrusion detection are essential for providing security. However, designing an intrusion detection method that is both efficient and lightweight enough for deployment on RSU devices remains an urgent challenge. To settle these issues, this paper proposes a novel intrusion detection rule generation method that optimise LLM‐Generated Intrusion Detection Rules with support vector machine (SVM). This approach automates the generation of intrusion detection rules for new attack types and employs SVM to optimise the parameter values within these rules, making the LLM‐generated rules more accurate. The generated detection rules are designed to achieve effective intrusion detection on most edge devices, providing a robust solution for enhancing the security of the RSU network. After the verification of the real machine experiment, the proposed IDS has a detection accuracy of 98.89% for the processed attack types.

Digital Twin (DT) technology has gained significant attention for simulating and optimizing urban traffic systems, especially in intersection vehicle trajectory prediction. However, digital twin traffic system faces significant challenges due to privacy and security regulations that prevent the centralized storage of trajectory and semantic data, which are essential for training accurate predictive models using sensitive traffic information. To address these issues, we introduce the integration of federated learning spatio-temporal-semantic attention-based trajectory (FedSTAST) model into the DT framework for vehicle trajectory prediction. In our FedSTAST, edge servers in physical space utilize local sensor data to perform computations and train models without transmitting raw data, only the model parameters are sent to a cloud server in the twin space for aggregation. This decentralized approach ensures data privacy while enabling collaborative model training. The simulation results demonstrate that the FedSTAST model effectively handles co-training and multi-source semantic input processing within spatio-temporal-semantic attention-based trajectory (STSAT) models, enhancing trajectory prediction accuracy and robustness in the dynamic, real-time context of DT-based urban traffic systems.

ABSTRACT An emerging component of smart cities is vehicle‐to‐grid (V2G) technology, which provides a novel approach to scheduling and energy storage. Security threats currently impede V2G's normal operations. V2G security faces two challenges. Current V2G security schemes only consider the static security approach, which is insufficient to handle the problem of advanced persistent attacks and high dynamics in V2G. However, the lack of a unified information modeling technique in present V2G causes problems with security and communication. The aim is to propose a novel technique in task allocation and security analysis based on smart energy management using a machine learning model in V2G architecture. Here, the smart energy management and task allocation are carried out using a hybrid fuel cell model with a deep vector Q‐gradient model. Then, the security analysis of the V2G network is carried out using a multilayer blockchain smart contract‐based federated LSTM model. Experimental analysis is carried out in terms of QoS, energy efficiency, network efficiency, data integrity, and training accuracy. Simulation results are conducted to prove the effectiveness of this proposed method.

Abstract Identifying malicious intent within a program, also known as malware, is a critical security task. Many detection systems remain ineffective due to the persistent emergence of zero‐day variants, despite the pervasive use of antivirus tools for malware detection. The application of generative AI in the realm of malware visualization, particularly when binaries are depicted as colour visuals, represents a significant advancement over traditional machine‐learning approaches. Generative AI generates various samples, minimizing the need for specialized knowledge and time‐consuming analysis, hence boosting zero‐day attack detection and mitigation. This paper introduces the Deep Convolutional Generative Adversarial Network for Zero‐Shot Learning (DCGAN‐ZSL), leveraging transfer learning and generative adversarial examples for efficient malware classification. First, a normalization method is proposed, resizing malicious images to 128 × 128 or 300 × 300 for standardized input, enhancing feature transformation for improved malware pattern recognition. Second, greyscale representations are converted into colour images to augment feature extraction, providing a richer input for enhanced model performance in malware classification. Third, a novel DCGAN with progressive training improves model stability, mode collapse, and image quality, thus advancing generative model training. We apply the Attention ResNet‐based transfer learning method to extract texture features from generated samples, which increases security evaluation performance. Finally, the ZSL for zero‐day malware presents a novel method for identifying previously unknown threats, indicating a significant advancement in cybersecurity. The proposed approach is evaluated using two standard datasets, namely dumpware and malimg, achieving malware classification accuracies of 96.21% and 98.91%, respectively.

The way digital assets are exchanged is rapidly evolving with the emergence of the Metaverse. In particular, immersive and decentralized platforms are reshaping transaction dynamics by offering new levels of interactivity. However, with these innovations come substantial challenges, especially those related to ensuring secure, energy-efficient, and interoperable asset flows across domains. Motivated by these needs, this paper present a transaction framework that is designed to balance scalability and sustainability. Specifically, we combine the relay chain–parachain architecture with chameleon hash signatures (CHS) to construct a mechanism capable of handling cross-domain digital content exchange. The proposed design aims to support flexible asset handling while keeping computation and energy demands in check. Taking inspiration from Green Supply Chain Management (GSCM), the framework also reflects concerns about sustainable digital asset governance. To evaluate its practicality, we carried out experiments across various network conditions and transaction volumes. The evaluation results suggest that our framework maintains low latency and reasonable energy efficiency, all while ensuring secure and interoperable transactions in the Metaverse context. These findings point to the potential of the framework as a stable infrastructure for digital asset management at scale in future decentralized environments.

The critical importance of monitoring and recognizing human emotional states in healthcare has led to a surge in proposals for EEG-based multimodal emotion recognition in recent years. However, practical challenges arise in acquiring EEG signals in daily healthcare settings due to stringent data acquisition conditions, resulting in the issue of incomplete modalities. Existing studies have turned to knowledge distillation as a means to mitigate this problem by transferring knowledge from multimodal networks to unimodal ones. However, these methods are constrained by the use of a single teacher model to transfer integrated feature extraction knowledge, particularly concerning spatial and temporal features in EEG data. To address this limitation, we propose a multi-teacher knowledge distillation framework enhanced with a Large Language Model (LLM), aimed at facilitating effective feature learning in the student network by transferring knowledge of extracting integrated features. Specifically, we employ an LLM as the teacher for extracting temporal features and a graph convolutional neural network for extracting spatial features. To further enhance knowledge distillation, we introduce causal masking and a confidence indicator into the LLM to facilitate the transfer of the most discriminative features. Extensive testing on the DEAP and MAHNOB-HCI datasets demonstrates that our model outperforms existing methods in the modality-incomplete scenario. This study underscores the potential application of large models in this field.

Transportation Cyber-Physical System (T-CPS) is a pivotal technology for advancing Intelligent Transportation System, integrating physical transportation infrastructure with network technology and computational algorithms. This integration facilitates real-time road condition monitoring, accurate traffic forecasting, and efficient traffic management to reduce congestion and enhance safety. However, relying on public channels for data transmission in T-CPS exposes it to numerous security threats. Addressing these challenges, this paper proposes an intelligence blockchain-based lightweight authentication protocol to enhance the security and trustworthiness of Intelligent Transportation System. The protocol is designed to resist common attacks, such as capture attacks and insider privileged personnel attacks, ensuring secure vehicle communication. Through rigorous Real-Or-Random model formal proofs, the security properties of the protocol are validated. Furthermore, the efficiency of the protocol is demonstrated, showing its lightweight nature and security robustness. This work represents a significant step toward secure, reliable, and efficient communication in vehicular networks, paving the way for more robust T-CPS applications, including autonomous driving and real-time traffic management.

Vehicular Ad hoc NETwork (VANET) refers to an arbitrarily distributed network that is an integral subset of an Intelligent Transport System (ITS) and plays a vital role in providing convenient transportation, improving traffic safety, etc.VANETrealizes interactive communication through wireless medium among Vehicle to Vehicle (V2V) and Vehicle to Infrastructure (V2I). But, owing to the open/public communication nature of the wireless medium, it is often prone to different security threats. On the other hand, security is a significant aspect of theVANETframework. Thus, to address the security concerns, it is a common practice to establish an authentication and key agreement protocol among the communicating entities that ensures to provide the essential security requirements. However, after extensive literature survey, it is identified that all the existing authentication protocols are prone to different security threats. Moreover, the implementation of these protocols in real-world scenarios becomes impractical as it bears higher computation and/or communication overheads. Therefore, in this paper, we have proposed a trust-extended mutual authentication protocol for theVANETframework using a Puncturable Pseudorandom Function (PPRF). The informal and formal security verification of our protocol proves that it provides comparably higher security than the existing schemes. Further, our protocol is simulated using a well-known AVISPA simulation tool and the results show that our protocol is SAFE against replay and man-in-the-middle attacks. Additionally, in comparison to the existing literature our protocol provides higher computation and communication efficiency. Therefore, our protocol is reliable and secure for real-world implementation in theVANETframework.

Traditional insurance policy models involve cumbersome multiparty verification and processing, leading to a prolonged and time-consuming procedure resulting in claim leakage. The existing insurance and blockchain-based systems have no module specifically for electric vehicles to cover physical damage. This becomes particularly significant in electric vehicles (EVs), where insurance is essential due to the high cost of vehicle parts and the vehicles themselves. Electric vehicles with various sensors and IoT devices are susceptible to physical damage and attacks. To address these challenges and provide robust financial support to policyholders, an enhanced blockchain-based electric vehicle insurance policy (BE-VIP) is proposed to cover vehicle damages. BE-VIP leverages sensory and telemetry data from vehicle sensors, IoT devices, and drivers’ behavior for a more comprehensive analysis. However, the insecure nature of the public network in the internet of electric vehicles (IoEV) exposes it to various security threats and attacks. Recognizing this, BE-VIP emphasizes implementing a lightweight privacy-preserving and efficient authentication protocol to enhance network security. A secure driver-driving score (DDS) is proposed to reward and punish the vehicle based on driving behavioral data and easy insurance policy transfer from the previous owner to the current owner. To prevent fraudulent accidental claims, a YOLOv8 model-based damage detection model is combined with IPFS to create permanent evidence of an accident. The feasibility of the BE-VIP model is rigorously evaluated through a comprehensive analysis, considering factors such as computational complexity and gas consumption required for execution over the Ethereum blockchain network.