Connected Autonomous Electric Vehicles (CAEVs) integrate complex software architectures and advanced connectivity features that, while enhancing transportation safety and efficiency, substantially expand cyberattack surfaces. Traditional vulnerability assessment methodologies, including static code analysis and fuzz testing, exhibit fundamental limitations when applied to the intricate real-time systems characteristic of CAEVs, frequently overlooking interaction-driven vulnerabilities while producing elevated false-positive rates that compromise safety-critical operations. This research presents GenSecure-CAEV, a novel generative AI framework employing fine-tuned large language models for proactive vulnerability discovery within CAEV software components and communication protocols. The framework systematically integrates automotive domain knowledge with advanced AI techniques to analyse critical system artefacts, including electronic control unit firmware and vehicle-to-everything communication logs, targeting complex vulnerabilities such as buffer overflows and protocol-specific logic flaws. Evaluation through controlled case studies simulating realistic CAEV operational environments demonstrates robust detection capabilities, achieving macro-F1 scores of 96.3% (AUTOSAR) and 95.8% (CAN Bus) with improved recall and optimised false-positive rates compared to conventional methods outperforming static analysis (81.6%), graph neural networks (86.9%), and generative fuzzing (77.1%). The framework addresses emerging security challenges in software-defined vehicles while maintaining alignment with ISO/SAE 21,434 regulatory frameworks and providing scalable implementation strategies. GenSecure-CAEV advances intelligent transportation security by establishing an evidence-based methodology for vulnerability discovery that enhances the cyber resilience of next-generation mobility systems against evolving threat landscapes.

The privacy and security of cloud data have drawn much attention, leading to more data owners outsourcing encrypted data. However, the common practice of encryption can reduce data searchability. Semantic searchable encryption aims to support flexible queries over encrypted data and achieve efficient search while ensuring that search results match the user's search intent. Although semantic searchable encryption schemes have made progress, they still have limitations in properly balancing accuracy, efficiency, and security. In this paper, we propose a novel Context-Enhanced Semantic Searchable Encryption (CESSE) scheme to achieve accurate and highly efficient secure semantic search over encrypted cloud data. To achieve it, we first adopt a context-enhanced pre-trained model component to mine the relevance between queries and documents by contrastive learning and obtain context-enhanced vector representations to improve search accuracy. Then, to ensure privacy protection, we utilize an optimized asymmetric scalar-product-preserving encryption (optimized ASPE) algorithm to encrypt vectors before outsourcing to the cloud. Additionally, we construct the approximate nearest neighbor (ANN) index to accelerate vector searching. At last, we give a formal definition of security and theoretically prove the safety of our scheme under a more practical threat model. Extensive experiments demonstrate that the CESSE outperforms state-of-the-art baselines with better accuracy and efficiency.

Driven by the rising popularity of multi-replica backups for enhanced data reliability and availability in cloud storage, multi-replica data auditing, which guarantees that cloud service providers (CSPs) securely store all designated replicas, has become a prominent research area in cloud data security. However, existing multi-replica auditing schemes pose a notable challenge: users incur additional computation and communication overheads to generate and upload multiple replicas. This approach also diverges from the conventional multi-replica storage model, where users typically submit a single copy and the CSP handles replica creation. To address this issue, this paper presents a novel multi-replica data auditing scheme based on Trusted execution environments (TEEs), named TEEMRDA. TEEMRDA is designed to align with real-world practices where users upload a single copy, significantly reducing user burden. To guarantee reliable multi-replica storage, we propose a random mask-based strategy implemented by TEEs to securely generate a predetermined number of data copies. For auditability, we introduce an impartial dual authentication mechanism using replica and data block index-independent signatures, employing both online and offline procedures. This approach substantially reduces the TEE’s computation overhead in generating tags for multiple-replica blocks and enhances efficiency for third-party auditors conducting data audits. Finally, we conduct comprehensive security validation and performance comparison of TEEMRDA with state-of-the-art schemes. The results demonstrate that TEEMRDA achieves secure and efficient auditing for multi-replica data, outperforming existing schemes in terms of computation and communication overheads.

It is well known that both the label information and the local geometry structure information are very important for image data clustering and classification. However, nonnegative matrix factorization (NMF) and its variants do not fully utilize the information or only use one of them. This paper presents a graph regularized discriminative nonnegative matrix factorization (GDNMF) for image data clustering, in which the local geometrical structure and label information of the observed samples are thoroughly considered. In the objective function of NMF, two constraint terms are added to preserve the above information. One is a sparse graph, which is adaptively constructed to obtain the local geometrical structure information. The other is data label information, which is used to capture discriminative information of the original data. By using local and label information, the proposed regularized discriminative nonnegative matrix factorization indeed improves the discrimination power of matrix decomposition. In addition, the F-norm formulation based cost function of regularized discriminative nonnegative matrix factorization is given, and the update rules for the optimization function of regularized discriminative nonnegative matrix factorization are proved. The experiment results on several public image datasets demonstrate the effectiveness of GDNMF algorithm. The innovation of this paper lies in extending unsupervised NMF to semi-supervised case and adaptively capturing the local structure of data based on sparse graph. However, the proposed method does not take into account the challenges of multiview data processing.

Due to advancement in Information and Communications Technology (ICT) and Internet of Things (IoT), the Internet of Drones (IoD) can be employed in numerous applications, facilitating the daily lives of diverse users, including civilians and others. Wireless communication nature leads to an IoD environment to be vulnerable to various potential attack risks, such as data breaches, man-in-the-middle, impersonation, replay, and data leaking attacks. As a result, the security of the IoD environment becomes crucial. To safeguard the data and devices (such as IoT-enabled drones and servers) integral to IoD networks, a security solution is essential. It is imperative to implement targeted security measures, such as intrusion detection, access control, and authentication, in order to establish a security scheme that is both reliable and efficient. In this article, we mainly focus on developing a secure authentication and key management scheme that leverages blockchain technology. Most existing authentication techniques proposed in IoT and IoD environments are either inefficient in communication and computation, or they are insecure against various attacks. To mitigate these issues, this study proposes a secure blockchain-based authentication and key management scheme for IoD applications (in short BAKMM-IoD). The blockchain is applied here as a secure data storage purpose. After performing a detailed security analysis and formal security verification with the widely-recognized Scyther tool, the proposed BAKMM-IoD has exhibited resilience against different potential attacks. BAKMM-IoD also surpasses other contemporary existing schemes in terms of security and functionality features, including computational costs, and communication costs. Moreover, the blockchain simulation shows that the influence of the proposed BAKMM-IoD on critical performance metrics in real-world scenarios.

To achieve real-time transmission of extensive remote sensing data, the combination of space-air-ground integrated networks (SAGINs) and free-space optical (FSO) technology is considered a suitable solution. However, the strongly attenuated atmosphere environment that covers long spatial distances and frequent handover of user-to-data links seriously affect the reliability of FSO links. As a relay linking the satellite layer and the ground layer, the introduction of high altitude platform (HAP) shortens the ultra-long satellite-to-ground FSO links, largely mitigating unreliable tropospheric connections. In order to achieve the improvement of the reliability of the whole SAGIN system, this paper integrates the space, air and ground three-layer network through HAP networking (HAPN). Based on HAPN architecture, the reliability enhancement scheme for two-hop FSO links is further designed. The first step is to reduce the attenuation of atmosphere turbulence by adjusting the HAP zenith angle through distanced clustering and Tabo search algorithms. The second step is to reduce the link handover times by extending the time window flexibly through HAP networking. The results show that the proposed schemes based on HAPN improve the reliability of the whole process of data transmission.

To empower the management of smart meters’ demand load within a smart grid environment, this paper presents a Feed-forward Neural Network with ADaptive Evolutionary Learning Approach (ADELA). In this model, the load forecasting information is propagated via neurons of input and multiple hidden layers and the final estimated output is achieved with the help of the sigmoid activation function. An improved evolutionary algorithm is proposed for training and adjusting the interconnecting weights among the layers of the intended neural network. This model is capable of addressing the critical challenges of high volatility, uncertainty, missing smart meters data, and sudden upsurge and plunge in electricity demand. The proposed algorithm is able to learn the best suitable evolutionary operators from a given pool of operators and the probabilities associated with them. The proposed load forecasting approach is simulated over three real-world smart meter datasets, including the Australian Smart Grid Smart City project, the Irish Commission for Energy Regulation, and UMass Smart. The performance evaluation and comparison of the proposed approach with the existing state-of-the-art approaches revealed a relative improvement of up to 46.93%, 5.05%, and 2.20% in forecast accuracy over the Smart Grid Smart City, UMass Smart and the Irish Commission for Energy Regulation datasets, respectively.

The energy market encompasses the behavior of energy supply and trading within a platform system. By utilizing centralized or distributed trading, energy can be effectively managed and distributed across different regions, thereby achieving market equilibrium and satisfying both producers and consumers. Additionally, the energy market can address future production control and environmental concerns such as energy shortages and environmental deterioration. However, recent years have presented unprecedented challenges and difficulties for the development of the energy market. These challenges include regional energy imbalances, volatile energy pricing, high computing costs, and issues related to transaction information disclosure. These factors have hindered the smooth operation of the energy market. Researchers widely acknowledge that the security features of blockchain technology can enhance the efficiency of energy transactions and establish the fundamental stability and robustness of the energy market. This type of blockchain-enabled energy market is commonly referred to as an energy blockchain. Currently, there is a burgeoning amount of research in this field, encompassing algorithm design, framework construction, and practical application. It is crucial to organize and compare these research efforts to facilitate the further advancement of energy blockchain. This survey aims to comprehensively review the fundamental characteristics of blockchain and energy markets, highlighting the significant advantages of combining the two. Moreover, based on existing research outcomes, we will categorize and compare the current energy market research supported by blockchain in terms of algorithm design, market framework construction, and the policies and practical applications adopted by different countries. Finally, we will address current issues and propose potential future directions for improvement, to provide guidance for the practical implementation of blockchain in the energy market.

In this paper, we address the challenging source-free unsupervised domain adaptation (SFUDA) for pinhole-to-panoramic semantic segmentation, given only a pinhole image pre-trained model (i.e., source) and unlabeled panoramic images (i.e., target). Tackling this problem is non-trivial due to three critical challenges: 1) semantic mismatches from the distinct Field-of-View (FoV) between domains, 2) style discrepancies inherent in the UDA problem, and 3) inevitable distortion of the panoramic images. To tackle these problems, we propose 360SFUDA++ that effectively extracts knowledge from the source pinhole model with only unlabeled panoramic images and transfers the reliable knowledge to the target panoramic domain. Specifically, we first utilize Tangent Projection (TP) as it has less distortion and meanwhile slits the equirectangular projection (ERP) to patches with fixed FoV projection (FFP) to mimic the pinhole images. Both projections are shown effective in extracting knowledge from the source model. However, as the distinct projections make it less possible to directly transfer knowledge between domains, we then propose Reliable Panoramic Prototype Adaptation Module (RP2AM) to transfer knowledge at both prediction and prototype levels. RP$^2$AM selects the confident knowledge and integrates panoramic prototypes for reliable knowledge adaptation. Moreover, we introduce Cross-projection Dual Attention Module (CDAM), which better aligns the spatial and channel characteristics across projections at the feature level between domains. Both knowledge extraction and transfer processes are synchronously updated to reach the best performance. Extensive experiments on the synthetic and real-world benchmarks, including outdoor and indoor scenarios, demonstrate that our 360SFUDA++ achieves significantly better performance than prior SFUDA methods.

With the unprecedented prevalence of Industrial Internet of Things(IIoT)and 5G technology,various applications supported by industrial communication systems have generated exponentially increased processing tasks,which makes task assignment inefficient due to insufficient workers.In this paper,an Intelligent and Trustworthy task assignment method based on Trust and Social relations(ITTS)is proposed for scenarios with many tasks and few workers.Specifically,ITTS first makes initial assignments based on trust and social influences,thereby transforming the complex large-scale industrial task assignment of the platform into the small-scale task assignment for each worker.Then,an intelligent Q-decision mechanism based on workers'social relation is proposed,which adopts the first-exploration-then-utilization principle to allocate tasks.Only when a worker cannot cope with the assigned tasks,it initiates dynamic worker recruitment,thus effectively solving the worker shortage problem as well as the cold start issue.More importantly,we consider trust and security issues,and evaluate the trust and social circles of workers by accumulating task feedback,to provide the platform a reference for worker recruitment,thereby creating a high-quality worker pool.Finally,extensive simulations demonstrate ITTS outperforms two benchmark methods by increasing task completion rates by 56.49％-61.53％and profit by 42.34％-47.19％.

For rapid on-orbit forwarding of high-resolution remote sensing (RS) images, the combination of very low-Earth orbit (VLEO) and optical intersatellite links (OISLs) has recently emerged as a focus of nonterrestrial networks (NTNs). However, at the height of VLEO, not only does the attenuation factor of the optical signal-to-noise ratio grow exponentially, but the service with more than ten times bandwidth expansion, which adds a significant burden to the buffer queue. To address the above reliability and latency challenges, this article proposes a routing framework based on the characteristics of the VLEO scenario. Specifically, the OISL path in VLEO is modelled first. In order to reduce the severe impact of multihop on the routing performance, we perform dynamic group scheduling for OISL. Additionally, taking advantage of the fact that RS missions can be scheduled in advance, we adopt a "route first, then establish links" approach to specifically plan service paths. Considering the coordination of a single service and overall network performance, this problem is solved using a Q-value decomposed multiagent reinforcement learning method. Simulation results demonstrate that our scheme maintains excellent reliability and latency performance under VLEO scenarios with varying heights, network scales, and traffic loads.

Device-free localization (DFL) represents an emerging technology in autonomous driving assistance systems, sensing targets around a vehicle without requiring them carrying any devices. Current machine learning methodologies for DFL struggle with adaptively extracting discriminative features, resulting in weak robustness. Although dictionary learning techniques have shown potentiality in robust feature extraction for images and signals, their direct application to improving the robustness of DFL presents significant challenges. Specifically, the unsupervised mechanism of dictionary learning complicates the calibration of target location bases. The complex optimization processes also lead to inefficient parameter updating. To address these issues, we propose a deep-supervised dictionary learning (DSDL) approach for enhancing the robustness of DFL. The proposed DSDL method synergizes the advantages of sparse representation and deep learning, incorporating the robustness and interpretability of dictionary learning with the efficient parameter updating characteristic of deep learning. The supervised mechanism enables collaborative labeling of the learned signal bases. Experimental results, derived from an established DFL system, demonstrate that our DSDL outperforms existing techniques in both robustness and accuracy. DSDL maintains strong robustness and achieves high localization accuracy, surpassing 99% on clean data and sustaining over 97% accuracy under 30 dB signal-to-noise ratio (SNR) conditions. This work highlights the potential of existing WiFi infrastructure to provide cost-efficiency solutions for target localization, paving the way for future applications of wireless sensing.

Unsupervised visible-infrared person re-identification (UVI-ReID) has recently gained great attention due to its potential for enhancing human detection in diverse environments without labeling. Previous methods utilize intra-modality clustering and cross-modality feature matching to achieve UVI-ReID. However, there exist two challenges: 1) noisy pseudo labels might be generated in the clustering process, and 2) the cross-modality feature alignment via matching the marginal distribution of visible and infrared modalities may misalign the different identities from two modalities. In this paper, we first conduct a theoretic analysis where an interpretable generalization upper bound is introduced. Based on the analysis, we then propose a novel unsupervised cross-modality person re-identification framework (PRAISE). Specifically, to address the first challenge, we propose a pseudo-label correction strategy that utilizes a Beta Mixture Model to predict the probability of mis-clustering based network's memory effect and rectifies the correspondence by adding a perceptual term to contrastive learning. Next, we introduce a modality-level alignment strategy that generates paired visible-infrared latent features and reduces the modality gap by aligning the labeling function of visible and infrared features to learn identity discriminative and modality-invariant features. Experimental results on two benchmark datasets demonstrate that our method achieves state-of-the-art performance than the unsupervised visible-ReID methods.

Image classification is a fundamental task in deep learning, and recent advances in quantum computing have generated significant interest in quantum neural networks. Traditionally, Convolutional Neural Networks (CNNs) are employed to extract image features, while Multilayer Perceptrons (MLPs) handle decision making. However, parameterized quantum circuits offer the potential to capture complex image features and define sophisticated decision boundaries. In this paper, we present a novel Hybrid Quantum–Classical Neural Network (H-QNN) for image classification, and demonstrate its effectiveness using the MNIST dataset. Our model combines quantum computing with classical supervised learning to enhance classification accuracy and computational efficiency. In this study, we detail the architecture of the H-QNN, emphasizing its capability in feature learning and image classification. Experimental results demonstrate that the proposed H-QNN model outperforms conventional deep learning methods in various training scenarios, showcasing its effectiveness in high-dimensional image classification tasks. Additionally, we explore the broader applicability of hybrid quantum–classical approaches in other domains. Our findings contribute to the growing body of work in quantum machine learning, and underscore the potential of quantum-enhanced models for image recognition and classification.

Threshold switching (TS) memristors are promising candidates for artificial neurons in neuromorphic systems. However, they often lack biological plausibility, typically functioning solely in an excitation mode. The absence of an inhibitory mode limits neurons' ability to synergistically process both excitatory and inhibitory synaptic signals. To address this limitation, we propose a novel memristive neuron capable of operating in both excitation and inhibition modes. The memristor's threshold voltage can be reversibly tuned using voltages of different polarities because of its bipolar TS behavior, enabling the device to function as an electronically reconfigurable bi-mode neuron. A variety of neuronal activities such as all-or-nothing behavior and tunable firing probability are mimicked under both excitatory and inhibitory stimuli. Furthermore, we develop a self-adaptive neuromorphic vision sensor based on bi-mode neurons, demonstrating effective object recognition in varied lighting conditions. Thus, our bi-mode neuron offers a versatile platform for constructing neuromorphic systems with rich functionality.

Some attackers in the Internet of Things submit falsified high-quality data to cause harm to users. To prevent malicious workers from reporting untruthful data for skyline computation in Mobile Crowd Sensing, we propose a double trust check-based spatiotemporal data acquisition scheme, DTC-MDD. In DTC-MDD, worker trust uses four-way validation to obtain reliable worker trust evaluations. Then, based on Probabilistic Skyline Calculation, we propose a worker selection algorithm to select high-trust, high-quality workers for data reporting. We also introduce the Non-Interactive Encrypted Integer Comparison Protocol to safeguard privacy between workers and users from malicious attacks. Finally, through extensive simulations on real datasets, DTC-MDD effectively enhances the quality and security of spatiotemporal data acquisition. DTC-MDD improved the data quality and reliability of candidate worker sets by 16.2% and 49.1%, respectively, and the data quality and reliability of the first skyline worker by 21.4% and 320.0%, respectively.

Quantum computing and machine learning (ML) have received significant developments which have set the stage for the next frontier of creative work and usefulness. This paper aims at reviewing various data-encoding techniques in Quantum Machine Learning (QML) while highlighting their significance in transforming classical data into quantum systems. We analyze basis, amplitude, angle, and other high-level encodings in depth to demonstrate how various strategies affect encoding improvements in quantum algorithms. However, they identify major problems with encoding in the framework of QML, including scalability, computational burden, and noise. Future directions for research outline these challenges, aiming to enhance the excellence of encoding techniques in the constantly evolving quantum technology setting. This review shall enable the researcher to gain an enhanced understanding of data encoding in QML, and it also suggests solutions to the current limitations in this area.

Digital twin (DT) technology uses Internet of Things (IoT) devices to collect real-world data and build a virtual world in the DT cloud. However, many IoT devices collect data in harsh natural environments, and these data cannot be transmitted through fixed base stations. Thus, many DT services adopt dynamic data transmission methods, such as transmission through unmanned aerial vehicle base stations (UAV-BSs). However, UAV-BS approaches have many communication constraints, such as limitations on the transmission bandwidth, data throughput, and number of channels. In addition, when integrating a large amount of data submitted by IoT devices, DT service providers need a corresponding mechanism to select the most valuable device data, which can be described by a winner decision problem with the goal of maximizing utility. In this paper, we consider the problem of maximizing the utility of a DT model under UAV-BS network transmission, transform it into a mixed integer programming model with communication and computing constraints, and adopt a reverse auction mechanism to solve it. Specifically, we design an optimal reverse auction mechanism based on optimal allocation and Vickrey–Clarke–Groves (VCG) theory. Additionally, a reverse auction mechanism with polynomial execution time is designed based on monotonic allocation, network maximum flow and critical value theory. These two mechanisms are proven to satisfy individual rationality and truthfulness. Experimental results indicate the favorable performance of the designed mechanisms.

The rapid advancement of internet technology has raised attention to the importance of electronic voting in maintaining democracy and fairness in elections. E-voting refers to the use of electronic technology to facilitate the casting and counting of votes in elections. The need for designated verification arises from concerns about voter privacy, auditability, and the prevention of manipulation. Traditional e-voting systems use cryptographic techniques for security but lack verifiable proof of integrity. Integrating e-voting with a quantum designated verifier could address these challenges by leveraging the principles of quantum mechanics to enhance security and trustworthiness. In light of this, we propose a quantum e-voting scheme that uses a designated verifier signature. To ensure the confidentiality and authenticity of the voting process, the scheme uses quantum features like the no-cloning theorem and quantum key distribution. The proposed scheme has security properties like source hiding, non-transferability, and message anonymity. The proposed scheme is resistant to many quantum attacks, such as eavesdropping and impersonation. Due to designated verification, the scheme minimizes the risk of tempering. This paper provides a detailed description of the proposed scheme and analyzes its security properties. Therefore, the proposed scheme is efficient, practical, and secure.

The current audio single-mode self-supervised classification mainly adopts a strategy based on audio spectrum reconstruction. Overall, its self-supervised approach is relatively single and cannot fully mine key semantic information in the time and frequency domains. In this regard, this article proposes a self-supervised method combined with knowledge distillation to further improve the performance of audio classification tasks. Firstly, considering the particularity of the two-dimensional audio spectrum, both self-supervised strategy construction is carried out in a single dimension in the time and frequency domains, and self-supervised construction is carried out in the joint dimension of time and frequency. Effectively learn audio spectrum details and key discriminative information through information reconstruction, comparative learning, and other methods. Secondly, in terms of feature self-supervision, two learning strategies for teacher-student models are constructed, which are internal to the model and based on knowledge distillation. Fitting the teacher’s model feature expression ability, further enhances the generalization of audio classification. Comparative experiments were conducted using the AudioSet dataset, ESC50 dataset, and VGGSound dataset. The results showed that the algorithm proposed in this paper has a 0.5% to 1.3% improvement in recognition accuracy compared to the optimal method based on audio single mode.