The rapid proliferation of Internet of Things (IoT) devices, coupled with their inherent security vulnerabilities, has significantly expanded the attack surface, intensifying threats such as man-in-the-middle attacks, traffic hijacking, and distributed denial-of-service (DDoS) attacks, thereby posing serious risks to the security and reliability of the entire ecosystem. The network traffic associated with IoT devices is diverse and dynamic, often exhibiting complex structural features such as periodic fluctuations, varying packet sizes, and time-varying patterns that make detection challenging. Although deep learning has demonstrated strong capabilities in efficiently identifying complex and dynamic malicious traffic through powerful feature extraction and adaptive learning abilities, its high model complexity, substantial computational demands, and large parameter sizes hinder direct deployment on resource-constrained IoT devices. In order to tackle this issue, this paper proposes a malicious traffic detection framework for the Internet of Things (IoT) based on multi-teacher knowledge distillation. The proposed model, termed the Lightweight Multi-Teacher Spatiotemporal Distillation Neural Network (LMT-SDNN), employs two high-performance teacher models: Residual Inception and a One-Dimension Convolution Netural Network (1D-CNN) integrated with idirectional Long Short-Term Memory (BiLSTM), to effectively capture the complex structural features of network traffic. Furthermore, a novel Time-Related Window Loss (TRW) function is design to enhance the student’s ability to capture temporal features, thereby improving its overall performance. The effectiveness of LMT-SDNN is validated through comparisons with five baseline models on two publicly available datasets, ToN_IoT and BoT_IoT. Experimental results show that LMT-SDNN achieves a compression rate of over 99% in both model complexity and parameter count, while maintaining an accuracy exceeding 99%, indicating its strong potential for multiclass malicious traffic detection in IoT environments.

The application of large language models(LLMs) in affective computing-ranging from empathetic chatbots to creative writing-has intensified the demand for distinguishing and authenticating AI-generated emotional content. Watermarking, by embedding detectable signals into the outputs of language models, offers a promising solution. However, a critical challenge persists: emotional texts often exhibit low entropy or complex spiky entropy distributions, which severely undermine the performance of existing watermarking methods. Unlike prior works that primarily treated spiky entropy as an external metric, we focus specifically on its role within the text generation process itself. To address the challenges of watermarking under low-entropy and complex entropy distributions, we propose DBW (Dynamic Bias Watermarking)-an entropy-aware watermarking algorithm for LLMs. DBW dynamically adjusts the watermarking bias in real time based on the entropy of each token. This innovation ensures a stronger watermark signal (increased green token count) in high-entropy contexts, while minimizing interference and quality degradation in fragile low-entropy emotional segments. Experimental results demonstrate that the proposed DBW algorithm outperforms the KGW watermarking method in both complex entropy distribution and low-entropy text generation scenarios. DBW achieves higher detection accuracy without sacrificing text quality. Furthermore, comparative experiments show that our proposed DBW algorithm demonstrates superior robustness under different attacks. Our work provides a reliable and adaptive tool for safeguarding emotion-AI generated content, contributing to the secure and trustworthy deployment of large-scale pre-trained models in affective computing.

In Vehicular Ad-hoc Networks (VANETs), vehicles exchange information in real-time by collaborating with Roadside Units (RSUs) and On-Board Units (OBUs) to enhance traffic efficiency and safety. Therefore, guaranteeing the integrity and confidentiality of messages, as well as the authenticity of the sender's identity, is crucial for establishing reliable communication. To address the aforementioned issues, this paper proposes an improved unforgeable aggregate signcryption scheme for VANETs, where a receiver can efficiently verify the legitimacy of multiple message sources simultaneously. The scheme enhances the security and privacy of VANETs by deploying a mapping between anonymous identities and public keys on a blockchain. Security analysis demonstrated that, under the hardness assumptions of the Elliptic Curve Discrete Logarithm Problem (ECDLP) and the Computational Diffie-Hellman Problem (CDHP), this scheme can achieve ciphertext unforgeability and confidentiality. Compared to existing related schemes, our proposed solution significantly reduces computational costs in both the signcryption and unsigncryption (especially in batch processing) phases.

The rise of multivariate time series (MTS) data has made prediction crucial, with deep learning models dominant yet vulnerable to adversarial attacks. These attacks use small perturbations on inputs to cause mispredictions. MTS data's inherent complexity and sensitivity demand stringent perturbation handling. To address this, we propose TFCA, an adversarial attack method focusing on feature dimension impact. TFCA calculates each feature's gradient contribution to predictions independently, establishing a feature importance ranking. It enhances temporal characteristics by integrating multi-dimensional elements (e.g., volatility, cosine direction) to constrain adversarial sample realism. Guided by this ranking, TFCA targets specific attack ratios on subsets of the time series, ensuring attack effectiveness while reducing perturbation size and improving stealthiness. Experiments on real MTS datasets against models (TCN, LSTNet, CNN) demonstrate TFCA's effectiveness, stealthiness, applicability, and transferability. The integration of post hoc explainable AI algorithms also provides interpretability. This research offers insights for enhancing MTS model robustness and security in practice.

Graph neural networks (GNNs) with differential privacy (DP) offer a reliable solution for safeguarding sensitive information within graph data. Nonetheless, existing DP-based privacy-preserving GNN learning frameworks generally overlook the local topological heterogeneity of graph nodes and tailor the same privacy budget for all nodes, which may lead to either overprotection or underprotection of some nodes, potentially diminishing model utility or posing privacy leakage risks. To address this issue, we propose a Topology-aware Differential Privacy Graph Neural Network learning framework, termed TDP-GNN, which can achieve personalized privacy protection for each node with improved privacy-utility guarantees. Specifically, TDP-GNN first identifies the topological importance of each node via an adjacency information entropy method. Then, the personalized topology-aware privacy budget is designed to quantify the privacy sensitivity of each node and adaptively allocate the privacy protection strength. Besides, a weighted neighborhood aggregation mechanism is proposed during the message-passing process of GNN training, which can eliminate the impact of the introduced differentiated DP noise on the utility of the GNN model. Since TDP-GNN is based on node-level local DP, it can be seamlessly integrated into any GNN architecture in a plug-and-play manner while ensuring formal privacy guarantees. Theoretical analysis indicates that TDP-GNN achieves ε-differential privacy over the entire graph nodes while providing personalized privacy protection. Extensive experiments demonstrate that TDP-GNN consistently yields better utilities when applied to various GNN architectures (e.g., GCN and GraphSAGE) across a diverse set of benchmarks.

Abstract To better provide fast computing services, vehicular edge computing can improve the quality of service and quality of experience for intelligent transportation in 6G by reducing task transmission delay. However, vehicular edge networks face network capability limitations and privacy issues in practice. High‐speed vehicles and the time‐varying environment make them unpredictable. In the meantime, smart vehicles with distinct computation capabilities need to process various tasks with different resource requirements, which will inevitably cause untimely task offloading and massive energy consumption. This paper proposes to use the space‐air‐ground integrated network with blockchain to enhance the network capability and the privacy protection of vehicular edge networks. The digital twin is taken to better capture the dynamic characteristics of vehicles and the entire environment. The urgency level is introduced to meet the delay requirements of different tasks, while considering the impact of digital twin deviation on task offloading. Moreover, the selection algorithm and the task distribution algorithm based on the improved genetic algorithmare are proposed to obtain the optimal offloading strategy. Simulation results demonstrate that, compared with the existing algorithms, the proposed scheme can maximize the system utility while diminishing the total time for task processing.

The rapid development of Internet of Things (IoT) technology has transformed people’s way of life and has a profound impact on both production and daily activities. However, with the rapid advancement of IoT technology, the security of IoT devices has become an unavoidable issue in both research and applications. Although some efforts have been made to detect or mitigate IoT security vulnerabilities, they often struggle to adapt to the complexity of IoT environments, especially when dealing with dynamic security scenarios. How to automatically, efficiently, and accurately understand these vulnerabilities remains a challenge. To address this, we propose an IoT security assistant driven by a Large Language Model (LLM), which, through the ICoT process, enhances the LLM’s understanding of IoT security vulnerabilities and related threats. The ICoT method we propose aims to enable the LLM to understand security issues by breaking down the various dimensions of security vulnerabilities and generating responses tailored to the user’s specific needs and expertise level. By incorporating ICoT, LLM can gradually analyze and reason through complex security scenarios, resulting in more accurate, in-depth, and personalized security recommendations and solutions. Experimental results show that, compared to methods relying solely on LLMs, our proposed LLM-driven IoT security assistant significantly improves the understanding of IoT security issues and provides personalized solutions based on user identity through the ICoT approach. From the evaluator’s perspective, it performs better across five dimensions.

360-degree panoramic video provides users with an unprecedented immersive experience and is rapidly evolving with the support of virtualized devices. Effective viewport prediction is crucial for alleviating high-speed bandwidth constraints and enhancing user quality of service. However, most existing research relies on saliency maps derived from multi-user viewport trajectories, often neglecting the rich multi-scale information inherent in individual user viewport trajectory. Inspired by signal decomposition theory, we propose an Empirical Mode Decomposition-based LSTM-MLP (EMD-ML) model. The EMD-ML extracts robust spatiotemporal representations from user viewport trajectory at multiple scales. Leveraging multi-scale representations and a hybrid learning framework, the model achieves accurate long-term viewport prediction from limited short-term viewport trajectory. The EMD-ML model achieves nearly a 50% reduction in orthodromic distance compared to state-of-the-art methods across four publicly available datasets. Additionally, experiments on different video types and training datasets show that our method generalizes well and can be applied in real-world scenarios.

Federated Learning, as a distributed learning model, enables multiple clients to collaboratively train models while preserving data privacy. However, recent studies have highlighted a potential drawback: sharing gradient information could unintentionally lead to the exposure of private training data, allowing attackers to reconstruct this data from the shared gradients. To defend against this threat while maintaining high model accuracy, we propose a defensive method called Gradient Perturbation-based Dynamic Clipping (GPDC). This method mitigates the risk of gradient information leakage by introducing minor perturbations to the shared gradients and employing dynamic clipping techniques to preserve model accuracy. It combines two approaches: gradient perturbation and segmented clipping. Clients use an adaptive mechanism to dynamically trim gradients. After trimming, noise scales are adaptively added, with the noise determined by the clipping threshold and gradient changes. The effectiveness of the proposed defensive strategy was evaluated through experiments on the MNIST and CIFAR10 datasets. The results reveal that the GPDC method successfully resists DLG attacks while maintaining high model performance.

Edge computing serves as a potent solution for distributed learning tasks, offering benefits such as low latency, reduced network load, enhanced data security and privacy, adaptability to IoT development, and flexible deployment options. Introducing greater randomness and stability through the random selection of clients for communication in each iteration round enhances overall system performance. To solve the problem of high communication cost in traditional federated learning and model performance degradation in heterogeneous environments, we propose a lightweight decentralized federated learning framework, which uses soft threshold ternary quantization to compress local models. Soft threshold ternary quantization is a method to compress weight parameters into discrete values, which significantly reduces the complexity of parameter representation, thus saving storage space and communication costs. We devise an algorithm for selecting clients with similar data distributions to enhance model accuracy and expedite convergence in heterogeneous environments. Our framework ensures model performance while minimizing communication costs and safeguarding user privacy. Experiments show that compared with the existing ternary quantization federated learning method, the accuracy of the model we trained is improved.

With integrity and traceability, blockchains have been widely applied in Internet of Things (IoT) systems. However, immutable blockchains contradict recent data regulations (e.g., the right to be forgotten in General Data Protection Regulation), making redactable blockchain-based IoT emerge as a promising paradigm. In this paradigm, IoT users can specify expressive policies (i.e., containing multiple logical AND and OR operators) to achieve controllable data editability. Unfortunately, existing related schemes with expressive policies face several issues: high communication costs, data privacy leakage (i.e., data can be read by all users), and inefficient user revocation. This article proposes a privacy-preserving and revocable redactable blockchain scheme in IoT systems, named BlockENC. BlockENC allows owners to specify expressive policies for controlling which users can read or edit their data and ensures downward compatible privileges (i.e., editable users own the privilege of readable users but not vice versa) under only$\mathcal {O}(n)$communication costs$(\mathcal {O}(n^{2})$in other schemes). The punchline of BlockENC is to define readability policies as subsets of editability policies and introduce access control trees to embed these policies in distributing data decryption keys and chameleon hash trapdoors. Moreover, drawing inspiration from ciphertext division mechanisms in proxy re-encryption techniques, BlockENC creates globally unique random values to reconstruct user keys, converting updating all existing keys or ciphertexts when user revocation cases occur into simply invalidating corresponding keys. Security analysis proves that BlockENC is secure against chosen-plaintext attacks. Experiments on the FISCO blockchain platform show that BlockENC achieves around$5\times $computation and$10\times $communication improvement over related works.

Federated learning is a distributed machine learning approach that enables multiple participants to collaboratively train a model without sharing their data, thus preserving privacy. However, the decentralized nature of federated learning also makes it susceptible to backdoor attacks, where malicious participants can embed hidden vulnerabilities within the model. Addressing these threats efficiently and effectively is crucial, especially given the impracticality of iterative and resource-intensive detection methods in federated learning environments. This article presents a novel framework for one-shot backdoor removal in federated learning. Our approach integrates advanced anomaly detection techniques with a unique model update aggregation strategy, allowing for the identification and neutralization of backdoor influences in a single update cycle without the need for extensive data access or communication between participants. Extensive experiments across various federated architectures and data distributions demonstrate that our method effectively mitigates backdoor threats while maintaining model performance and scalability. This work not only enhances the security of federated models but also contributes to the broader applicability of federated learning in sensitive and critical domains.

Owing to its practical configuration to edge computing and privacy preservation capabilities, federated learning (FL) has been increasingly appealing in Internet of Things (IoT) networks. However, due to the inherent openness of IoT network architectures, FL clients are susceptible to various attacks, resulting in unreliable local model updates. To address this challenge, federated unlearning (FU) emerges as a viable solution, which can erase such unreliable updates from the FL model using the unlearning operation while preserving model accuracy. Existing FU studies have significant potential, but they are not directly applicable to IoT networks because of their high computational costs and limited capacity to defend against prevalent dynamic attacks in mobile network environments. In this work, we propose FedRemover, a novel FU method specifically tailored for deployment in IoT networks. The key insight behind FedRemover is that model updates will exhibit inconsistency when exposed to attacks. Therefore, we devise a real-time malicious client detection scheme by examining the performance consistency of model updates. Upon detecting malicious clients, FedRemover promptly executes the unlearning operation, achieving an unlearned global model within a minimal number of rounds. This makes FedRemover highly efficient and robust against dynamic attacks, enabling it well-suited for practical deployment in IoT networks. Experiments on three standard datasets demonstrate the efficiency and robustness of FedRemover, with an obvious speed-up of 10× and comparable robustness guarantees compared with benchmark algorithms.

In unmanned aerial vehicle networks (UAVNs), unmanned aerial vehicles with restricted computing and communication capabilities can perform tasks in collaborative manner. However, communications in UAVN confront many security issues, for example, malicious entities may launch impersonate attacks. In UAVN, the command center (CMC) needs to perform mutual authentication with unmanned aerial vehicles in clusters. The aggregator (AGT) can verify the authenticity of authentication request from CMC; then, the attested authentication request is broadcasted to the reconnaissance unmanned aerial vehicle (UAV) in the same cluster. The authentication responses from UAVs can be verified and aggregated by AGT before being sent to CMC for validation. Also, existing solutions cannot resist malicious key generation center (KGC). To address these issues, this paper proposes a pairing-free authentication scheme (CLAS) for UAVNs based on the certificateless signature technology, which supports batch verification at both AGT and CMC sides so that the verification efficiency can be improved greatly. Security analysis shows that our CLAS scheme can guarantee the unforgeability for (attested) authentication request and (aggregate) responses in all phases. Performance analysis indicates that our CLAS scheme enjoys practical efficiency.

The fast penetration of Intelligent Connected Vehicles (ICVs) has become the primary growth engine of the automotive industry in recent years. Urban vehicular network consisting of ICVs is evolving towards a distributed intelligent platform for pervasive sensing, connecting and computing in Intelligent Transportation System (ITS) and smart cities. In this paper, we propose that parked vehicles (PVs) could be exploited for environment perception and model inference. We describe the system architecture and its typical application scenarios of distributed environment perception for city roads, parking lots, as well as for commercial and residential buildings. PVs are motivated to assist in deep learning model inference for the captured image data in such applications. Regarding the diversity of PVs in deep learning capability, a differential incentive mechanism is elaborately designed based on contract theory to emulate PVsparticipation. The experiment on the dataset of German Traffic Sign Recognition Benchmark is conducted to verify the effectiveness and efficiency of the proposed approach.

With the repaid development of Internet of Vehicles (IoV), more available resources and energy-efficient optimizations in resources scheduling are exactly required for large-scale network implementation for sustainable development. We observe that parked vehicles (PVs) have rich and underutilized resources for task execution. By scheduling them as general computing nodes to undertake computation tasks, we introduce a new computing paradigm, named by parked vehicular computing (PVC). There exists some challenging issues to be addressed for the facilitation of PVC. In particular, an incentive mechanism is needed to offer optimized rewards for PVs with the consideration of their parking time and energy consumption. In this paper, we investigate an energy-efficient PVC paradigm, and we design a contract-based incentive mechanism to motivate PVs to contribute their idle on-board resources. The PVs are classified into different types according to their parking time. Then, the designed contracts are assigned to different types of PVs. To realize the incentive mechanism, the optimization problem with the contract design is formulated to maximize the utility of the service provider. For optimal contract design, we solve the simplified problem by using Lagrangian multiplier method. Numerical results indicate that the proposed PVC with optimal contract design outperforms existing work in improving social welfare of resource scheduling, which takes quality-of-service and overall energy consumption into consideration. We also demonstrate that the contract-based incentive mechanism is energy-efficient and effective.