To overcome the challenges of trust isolation and scalability bottlenecks in cross-domain collaboration within the Industrial Internet of Things (IIoT), this paper proposes LBCM, a lightweight authentication protocol based on consortium blockchain and Chebyshev chaotic mapping. The protocol establishes a hybrid on-chain storage and off-chain verification trust architecture, and leverages the semi-group property of Chebyshev polynomials to achieve efficient and secure mutual authentication and key agreement. To address the storage limitations of blockchain, we design a global Merkle Tree commitment mechanism governed by smart contracts. Acting as consortium nodes, Trusted Authorities collaboratively maintain the global Merkle Root through a Propose-Vote-Execute consensus process, thus fixing the on-chain storage overhead at a constant level (O(1)), regardless of network scale. Furthermore, the protocol incorporates a “Filter-First, Access-Later” proactive defense strategy, enabling core nodes to rapidly filter out malicious traffic at the edge and effectively defending against Distributed Denial of Service (DDoS) attacks. Formal verification using ProVerif and comprehensive performance evaluations demonstrate that LBCM significantly outperforms existing mainstream schemes in terms of computational efficiency, communication overhead, and storage cost, while maintaining high security guarantees.

Malware Traffic Classification (MTC) is a key area of research in cybersecurity. Accurate and efficient classification is vital for defending against malware. Many methods use supervised learning, especially Convolutional Neural Networks (CNNs), to train feature extractors. However, obtaining a large number of labeled samples is costly, and relying solely on CNNs may limit the local receptive fields, compromising the retention of key features. Additionally, existing Transformer backbone networks have quadratic computational complexity. This paper proposes the MTC-Kansformer, a traffic classification model that addresses these issues and improves classification accuracy and efficiency. The model integrates a self-supervised learning framework with a Kansformer. The Kansformer encoder replaces the traditional Multi-Layer Perceptron (MLP) layer with fasterKAN, enhancing the representation and interpretability of nonlinear features. Initially, the raw traffic is converted into gray images. Then, the Kansformer self-supervised model is used to extract key features from randomly masked images, and the prediction module is employed to predict the masked images. Finally, the encoder is finetuned using a downstream malware dataset to perform effective malware traffic classification tasks. Experimental results show that the MTC-Kansformer outperforms existing models, achieving a classification accuracy of$\mathbf{9 8. 7 8 \%}$on the USTC-TFC2016 dataset, surpassing existing models.

With the evolution of edge intelligence technologies, federated learning has become a prominent decentralized learning framework, facilitating joint model training among numerous devices without compromising data privacy. However, federated learning confronts several challenges in practical applications, including data heterogeneity and edge device instability. To tackle these issues, this article introduces pFedMLKD, an innovative personalized federated learning framework that incorporates hierarchical knowledge distillation and leverages historical global model insights to enhance local model training, thereby reducing the effects of data heterogeneity. This method enhances learning efficiency and improves the generalization of models within federated learning systems. We establish theoretical guarantees for pFedMLKD’s optimization under nonconvex settings through rigorous convergence analysis. Empirical validation encompasses three benchmark datasets (CIFAR-10, CIFAR-100, and EMNIST) under strict nonindependent and identically distributed data partitioning protocols. The results demonstrate that our algorithm outperforms current methods in federated learning scenarios.

With the rapid development of Artificial Intelligence of Things (AIoT), the volume of generated data has surged exponentially. Deep compressive sensing (DCS) technology enables accurate data reconstruction at sub-Nyquist sampling rates, significantly enhancing data transmission efficiency and optimizing performance in unimodal vision tasks. However, given the substantial memory and computational resources required, deploying these Machine Learning(ML) models on resource-constrained edge devices and executing dynamic training poses significant challenges. Furthermore, cloud-based processing introduces privacy vulnerabilities and potential performance degradation. This paper introduces a ConvMixed-ViT architecture based on Low-Rank Adaptation (LoRA) and employs an end-to-end methodology integrating learnable CS to facilitate adaptive compression and precise reconstruction on edge environments. Specifically, low-rank adaptive frozen pretrained model weights are injected into the ConvMixed-ViT Architecture’s Transformer variant framework through trainable rank decomposition matrices, greatly reducing the number of trainable parameters for downstream tasks. The experimental results demonstrate that our proposed architecture achieves excellent reconstruction performance on standard benchmark datasets. In common resource-constrained scenarios, the model training scale is not only significantly reduced compared to other algorithms, while the performance is continuously optimized. This demonstrates the superior adaptability and learning efficiency of the model, in dynamic edge computing environments, providing strong support for future edge AI applications in various scenes, including driverless vehicles, smart cities, industrial manufacturing, healthcare services and so on.

The Internet of Medical Things (IoMT) is a promising new frontier in healthcare, enabling remote patient monitoring and telemedicine services. However, it also raises significant security and privacy concerns. To address these challenges, we propose a mutual authentication schema for IoMT using the Ring Learning with Errors (RLWE) and Boneh-Boyen signatures. Our scheme utilizes RLWE-based encryption and Boneh-Boyen signatures to ensure the confidentiality and integrity of data. We validate the authentication of the proposed scheme using Scyther verification and demonstrate its efficiency by performance evaluation. The results of performance evaluation demonstrate that our proposed scheme takes less computation and communication costs compared with other existing solutions. Therefore, it is a fairly attractive solution for efficient and secure mutual authentication in IoMT.

Wireless Sensor Networks (WSNs), a vital component of the Internet of Things (IoT), play an essential role in applications like environmental monitoring, smart homes, and industrial automation. For resource-limited WSNs, an effective data collection method can significantly reduce energy consumption and extend network life. Compressive sensing technology offers an efficient sampling method, significantly reducing data transmission volume in the network and allowing for original data restoration under certain noise levels. This is considered a promising approach. The core contribution of this paper is the development of an energy-efficient routing algorithm for WSNs, combining the principles of compressive sensing and ant colony algorithms. This novel algorithm reimagines data aggregation in clustered networks through compressive sensing, altering the paradigm of traditional intra-cluster communication. It introduces a novel intra-cluster routing optimization method and redesigns the ant colony algorithm to facilitate this optimization. Simulation results show that the proposed EARCS method significantly reduces overall network energy consumption and markedly extends network life compared to traditional methods.

The rapid development of 5 G and Internet of Things (IoT) has led to a surge in connected wireless devices, creating a significant demand for computation due to the evolution of intelligent applications. Addressing the computational pressure on Mobile Edge Computing (MEC) is now an urgent concern. Effectively offloading intensive tasks via Device-to-Device (D2D) links to the base station (BS) or idle devices improves computation quality and reduces latency. In this paper, we propose a D2D-assisted MEC system to address the scheduling challenges posed by numerous independent computing tasks generated by multiple users. We consider splitting the user’s task into multiple independent subtasks and calculating offloading separately to reduce processing delay. We represent this scheduling problem in the form of a task permutation and propose an improved Grey Wolf Optimizer (IGWO) metaheuristic algorithm to search for the optimal scheduling solution. This approach, through improvements to the nonlinear convergence factor and dynamic weighting, enhances the optimization speed and accuracy of the Grey Wolf algorithm, effectively reducing task processing latency. Simulation results indicate that the IGWO metaheuristic algorithm outperforms other benchmark methods in addressing this scheduling problem.

With the rapid development of 6G networks and artificial intelligence of things(AIoT), the volume of generated data has surged exponentially. Deep Compressed sensing(DCS) achieve accurate data reconstruction at Sub-Nyquist rates, minimizing data transmission volume and optimizing performance in unimodal vision tasks. However, given the substantial memory and computational resources required, deploying these DCS models on resource-constrained edge devices and executing dynamic training poses formidable challenges. Additionally, cloud-based processing introduces privacy vulnerabilities and potential performance degradation. This paper introduces a ConvMixed-ViT architecture based on Low-Rank Adaptation (LoRA) and employs an end-to-end methodology integrating learnable CS to facilitate compression and precise reconstruction on edge environments. Specifically, the LoRA freezes pre-trained model weights and injects them into the ConvMixed-ViT Architecture's Transformer variant framework through trainable rank decomposition matrices, greatly reducing the number of trainable parameters for downstream tasks and controlling the model size. The experimental results demonstrate that our proposed architecture achieves excellent reconfiguration performance on standard benchmark datasets. In resource-limited environments, CTLCS offers efficient adaptability and optimizes performance, proving its suitability for future edge AI applications.

Traditional machine learning involves collecting data from clients to a central server, where data may not be willingly shared by the clients. In contrast, federated learning (FL) trains a model without data sharing. The parameter server sends a global model to all clients, who train it locally and send local updates back. A major challenge is the high communication overhead from numerous local updates. To address this communication overhead, several algorithms have been proposed for FL tasks, such as sparsification and quantization. In this article, we propose a method called universal vector quantized compressive sensing for communication-efficient FL (FedUVeQCS). We demonstrate that combining universal vector quantized compressive sensing with FL enables dimensionality reduction and quantization of the local model updates without the need to consider the distribution of the reduced data. The quantization distortion caused by universal vector quantization can be considered as a negligible additive noise term. We evaluate FedUVeQCS on image classification tasks using MNIST and fashion-MNIST data sets and compare it with baseline algorithms, including FedAvg, universal vector quantization for FL, Top-k, and FedPAQ. Numerical results show the superiority of FedUVeQCS over baseline algorithms in terms of the number of bits uploaded while maintaining testing accuracy comparable to FedAvg.

To meet the low latency demands of numerous Internet of Things (IoT) applications, edge computing (EC) has been proposed to migrate computation from the cloud to the network's edge. This paper investigates the resource allocation problem in edge computing with soft deadline tasks and designs an efficient online bargaining scheme for resource allocation at edge nodes. We introduce task value functions to model the sensitivities of different tasks to latency. Then, we model the resource allocation problem as an online bargaining problem base on a three-layer edge computing model. We propose a specific two-period bargaining scheme. Corresponding pricing strategies are formulated for different periods to achieve optimal resource allocation and maximization of average utility(AU). Experimental results demonstrate that our algorithm outperforms other bargaining strategies and effectively improves system response speed.

In the federated learning (FL) paradigm, edge devices use local datasets to participate in machine learning model training, and servers are responsible for aggregating and maintaining public models. FL cannot only solve the bandwidth limitation problem of centralized training, but also protect data privacy. However, it is difficult for heterogeneous edge devices to obtain optimal learning performance due to limited computing and communication resources. Specifically, in each round of the global aggregation process by the FL, clients in a ‘strong group’ have a greater chance to contribute their own local training results, while those clients in a ‘weak group’ have a low opportunity to participate, resulting in a negative impact on the final training result. In this paper, we consider a federated learning multi-client selection (FL-MCS) problem, which is an NP-hard problem. To find the optimal solution, we model the FL global aggregation process for clients participation as a potential game. In this game, each client will selfishly decide whether to participate in the FL global aggregation process based on its efforts and rewards. By the potential game, we prove that the competition among clients eventually reaches a stationary state, i.e. the Nash equilibrium point. We also design a distributed heuristic FL multi-client selection algorithm to achieve the maximum reward for the client in a finite number of iterations. Extensive numerical experiments prove the effectiveness of the algorithm.