In beyond 5G multi-cell networks, cell-edge users generally experience poor communication quality due to their greater distance from the base station (BS), and increased interference from neighboring cells, which severely impacts their user experience. To address this issue, achieving fair and efficient resource scheduling is key to ensuring the quality of service for edge users. Therefore, this paper formulates a joint optimization problem of spectral subband selection and power control, aiming to maximize the proportional fair sum rate of the multi-cell network. However, most existing algorithms require instantaneous global channel state information, which results in poor scalability and is impractical, especially in highly dynamic wireless network environments with user mobility. Noting that the considered problem can be modeled as a decentralized partially observable Markov decision process, we propose a multi-agent deep reinforcement learning (MADRL) scheme based on a hierar chical centralized training and distributed execution framework (MADRL-HE), enabling agents to make spectral subband and transmit power selections using only local information and some outdated non-local information. Simulation results demonstrate that the proposed scheme features excellent scalability and fast convergence. Moreover, its proportional fair sum rate performance consistently outperforms the existing MADRL scheme in dynamic environments and surpasses centralized iterative optimization schemes in most dynamic scenarios.

Considering the harsh and complex environments in Underwater Wireless Sensor Networks (UWSNs), where the malicious nodes exist, the secure routing design is investigated in this paper. The malicious nodes are a serious threat to the security of sensor networks. This paper proposes a secure routing protocol, named Trust Based Routing (TBR), that focus on how to evaluate and find the malicious nodes and then determine the reliable routing. The core idea of TBR is that a new trust management model is designed by considering the various historical behaviors of nodes in the interaction process, which is able to assess the trust value of a node based on past behaviors between nodes and protect against potential attacks in the network. And the extra factors such as the remained energy of nodes, the distance between nodes and the trustiness are included in the routing criterions. Finally, a strategy for selecting relay nodes is proposed based on these considerations. Simulation results show that the proposed TBR algorithm can effectively defend against attacks from inside the network and is also more efficient and reliable compared to other existing routing protocols.

Secure data transmission and real-time state updates are critical yet challenging requirements for Cyber-Physical Systems (CPS) to maintain stability under adversarial conditions. While existing works have separately explored covert communications for security and Age of Information (AoI) optimization for timeliness, their interdependencies remain unaddressed, leading to suboptimal trade-offs between detection resistance, control performance, and resource efficiency. To bridge this gap, this paper proposes a novel co-design framework that jointly optimizes covert communication and AoI-aware control strategies. First, we rigorously derive a linear relationship between average AoI and control cost, termed the Cost of Update Delay (CoUD), which quantifies how outdated information exacerbates state fluctuations and increases stabilization efforts. Building on Kosta et al.’s Geo/Geo/1 queuing model, a closed-form expression for average AoI is further established as a function of sampling rate and packet delivery probability, explicitly linking communication parameters to control efficacy. Subsequently, this paper formulate a constrained optimization problem to minimize CoUD while guaranteeing covertness against eavesdroppers, leveraging Dinkelbach’s transformation and Lagrangian duality to decouple nonlinear constraints, derive optimal sampling rates, and transmission powers. Simulation results demonstrate that the proposed co-design framework achieves significant reductions in CoUD and superior freshness compared to baseline methods, while robustly maintaining covertness requirements. Notably, the proposed integration of sampling rate adaptation into the detection error rate model markedly enhances both AoI performance and resource efficiency, outperforming state-of-the-art disjoint designs. Therefore, this work provides a unified methodology to harmonize security, timeliness, and stability in resource-constrained CPS.

Increasing demands for Quality of Experience (QoE) lead to massive connectivity and intensive computation in future vehicular networks. This article proposes a device-to-device (D2D)-based mobile edge computing (MEC) network architecture to provide effective communication connections and sufficient computing abilities for vehicular networks. However, the available communication and computing resources are limited in the D2D-based vehicular MEC networks, and an imbalanced resource allocation always leads to suboptimal optimization of overall performances. To address this challenge, we formulate a Lyapunov optimization method-based resource allocation framework to balance communication and computing by compromising energy efficiency (EE) and time delay. However, the long-term resource allocation framework is ineffective when it ignores the dynamic characteristics of vehicular networks, i.e., channel state changes due to the movement of vehicles and a dynamic queue backlog with data fluctuations. Considering the time-varying channel state and dynamic queue backlog, the proposed framework aims to balance resource allocations while primarily maintaining network stability. Finally, we propose a Lyapunov optimization-based long-term dynamic resource allocation algorithm to develop real-time allocation strategies. Simulation results illustrate that the proposed algorithm balances communication and computing resources by tuning the control parameter V. Furthermore, the results confirm that the proposed algorithm outperforms baseline algorithms in real-time transmission and offloading ability.

Mobile-edge computing (MEC), pushing the centralized cloud computing, storage, and communication capability to the edge close to vehicular terminals, is proposed as a promising solution to support computation-intensive and delay-sensitive services. This article proposes a cooperative MEC-enabled task offloading framework where the computational task of each vehicle is divided and computed by multiple collaborative MECs located on the roadside. However, existing MEC-enabled offloading research is based on offline settings or static networks and fails to address the dynamic communication environments. These dynamic environments involve variations in temporality (real-time channel state) and spatiality (uncertain data-queue backlogs as vehicles pass through different coverage areas of MECs). In the dynamic vehicular networks, the degradation of utility energy efficiency (UEE) and time delay is inevitable and significantly impacted. To tackle this issue, we propose an online dynamic scheme to solve the problem of maximizing UEE while meeting time-delay constraints. We then introduce a novel online dynamic optimization algorithm based on Lyapunov optimization theory to adaptively create strategies for task offloading and communication resource allocation in parallel. Numerical simulations demonstrate that the proposed algorithm achieves a balance between UEE and delay, striking a flexible tradeoff by tuning the control parameter$V$. Furthermore, the results confirm that the proposed algorithm outperforms baseline algorithms in terms of real-time communication and transmission capability.

Summary While recent studies on automated multilabel chest X‐ray (CXR) images classification have shown remarkable progress in leveraging complicated network and attention mechanisms, the automated detection on chest radiographs is still challenging because the pathological patterns are usually highly diverse in their sizes and locations. The CNN model will suffer from the complicated background and high diversity of diseases, which reduce the generalization and performance of the model. To solve these problems, we propose a dual‐distribution consistency (DDC) model, which increases the consistency from two aspects, that is, feature‐level and label‐level. This model integrates two novel loss functions: multilabel response consistency (MRC) loss and distribution consistency (DC) loss. Specifically, we use the original image and its transformed image as inputs to imitate different views of CXR images. The MRC loss encourages the multilabel‐wise attention maps to be consistent between the original CXR image and its transformed counterpart. And the DC loss can force their output probability distributions to be uniform. In this manner, we can make sure that the model can learn discriminative features by using a different view of CXR images. Experiments conducted on the ChestX‐ray14 dataset show the effectiveness of the proposed method.

Recent research has revealed that the output of deep neural networks (DNNs) can be easily altered by adding relatively small perturbations to the input vector. In this paper, we analyze an attack in an extremely limited scenario where only one pixel can be modified. For that we propose a novel method for generating one-pixel adversarial perturbations based on differential evolution (DE). It requires less adversarial information (a black-box attack) and can fool more types of networks due to the inherent features of DE. The results show that 67.97% of the natural images in Kaggle CIFAR-10 test dataset and 16.04% of the ImageNet (ILSVRC 2012) test images can be perturbed to at least one target class by modifying just one pixel with 74.03% and 22.91% confidence on average. We also show the same vulnerability on the original CIFAR-10 dataset. Thus, the proposed attack explores a different take on adversarial machine learning in an extreme limited scenario, showing that current DNNs are also vulnerable to such low dimension attacks. Besides, we also illustrate an important application of DE (or broadly speaking, evolutionary computation) in the domain of adversarial machine learning: creating tools that can effectively generate low-cost adversarial attacks against neural networks for evaluating robustness.

The Internet of Things (IoT) is an extension of the traditional Internet, which allows a very large number of smart devices, such as home appliances, network cameras, sensors and controllers to connect to one another to share information and improve user experiences. IoT devices are micro-computers for domain-specific computations rather than traditional function-specific embedded devices. This opens the possibility of seeing many kinds of existing attacks, traditionally targeted at the Internet, also directed at IoT devices. As shown by recent events, such as the Mirai and Brickerbot botnets, DDoS attacks have become very common in IoT environments as these lack basic security monitoring and protection mechanisms. In this paper, we propose a novel light-weight approach for detecting DDos malware in IoT environments. We extract the malware images (i.e., a one-channel gray-scale image converted from a malware binary) and utilize a light-weight convolutional neural network for classifying their families. The experimental results show that the proposed system can achieve 94:0% accuracy for the classification of goodware and DDoS malware and 81:8% accuracy for the classification of goodware and two main malware families.