Internet of Agents (IoA) envisions a unified, agent-centric paradigm where heterogeneous large language model (LLM) agents can interconnect and collaborate at scale. Within this paradigm, federated fine-tuning (FFT) serves as a key enabler that allows distributed LLM agents to co-train an intelligent global LLM without centralizing local datasets. However, the FFT-enabled IoA systems remain vulnerable to model poisoning attacks, where adversaries can upload malicious updates to the server to degrade the performance of the aggregated global LLM. This paper proposes a graph representation-based model poisoning (GRMP) attack, which exploits overheard benign updates to construct a feature correlation graph and employs a variational graph autoencoder to capture structural dependencies and generate malicious updates. A novel attack algorithm is developed based on augmented Lagrangian and subgradient descent methods to optimize malicious updates that preserve benign-like statistics while embedding adversarial objectives. Experimental results show that the proposed GRMP attack can substantially decrease accuracy across different LLM models while remaining statistically consistent with benign updates, thereby evading detection by existing defense mechanisms and underscoring a severe threat to the ambitious IoA paradigm.

Future wireless networks must support emerging applications where environmental awareness is as critical as data transmission. Integrated Sensing and Communication (ISAC) enables this vision by allowing base stations (BSs) to allocate bandwidth and power to mobile users (MUs) for communications and cooperative sensing. However, this resource allocation is highly challenging due to:(i)dynamic resource demands from MUs and resource supply from BSs, and(ii)the selfishness of MUs and BSs. To address these challenges, existing solutions rely on either real-time (online) resource trading, which incurs high overhead and failures, or static long-term (offline) resource contracts, which lack flexibility. To overcome these limitations, we propose theFuture Resource Bank for ISAC, a hybrid trading framework that integrates offline and online resource allocation through a level-wise client model, where MUs and their coalitions negotiate with BSs. We introduce two mechanisms:(i)Offline Role-Friendly Win-Win Matching (offRFW2M), leveraging overbooking to establish risk-aware, stable contracts, and(ii)Online Effective Backup Win-Win Matching (onEBW2M), which dynamically reallocates unmet demand and surplus supply. We theoretically prove stability, individual rationality, and weak Pareto optimality of these mechanisms. Through comprehensive experiments, we show that our framework improves social welfare, latency, and energy efficiency compared to existing methods.

Future 6 G networks are envisaged to tightly integrate communication, sensing, and computing, demanding real-time, intent-driven intelligence at the edge. Whilelargelanguagemodels (LLMs) excel in intent recognition and semantic reasoning, their application to real-time network lifecycle management at the edge is limited by heterogeneousapplicationintents (APPIs), dynamic network conditions, and severe resource constraints. This paper proposes a novel lightweight LLM architecture, KGLlama-KD, that synergizes knowledge graphs (KGs) withknowledgedistillation (KD) to enable intent-driven networking and enhance 6 G edge intelligence. Specifically, a KG is constructed to formally describe the relationships among application scenarios, functional primitives, performance requirements within APPIs, and the correspondences between APPIs andnetworkservicerequests (NSRs), thereby producing a structured intent training dataset. Building upon the Llama 3 foundation model, a two-phase optimization framework is designed to support lightweight edge deployment while preserving translation fidelity. The LLM is first fine-tuned with KG guidance and compressed via KD in the cloud, and then deployed on resource-constrained edge nodes to perform real-time, accurate, and efficient APPIs interpretation. Experiments validate that KGLlama-KD achieves 95% accuracy for APPI understanding, surpassing DeepSeek and Qwen by an average of 8%. The distilled model reduces inference latency by 60% compared to full-scale LLMs, fulfilling the sub-100 ms requirement for 6 G latency-sensitive services.

EdgeIoT represents an approach that brings together mobile edge computing with Internet of Things (IoT) devices, allowing for data processing close to the data source. Sending source data to a server is bandwidth-intensive and may compromise privacy. Instead, federated learning allows each device to upload a shared machine-learning model update with locally processed data. However, this technique, which depends on aggregating model updates from various IoT devices, is vulnerable to attacks from malicious entities that may inject harmful data into the learning process. This paper introduces a new attack method targeting federated learning in EdgeIoT, known as data-independent model manipulation attack. This attack does not rely on training data from the IoT devices but instead uses an adversarial variational graph auto-encoder (AV-GAE) to create malicious model updates by analyzing benign model updates intercepted during communication. AV-GAE identifies and exploits structural relationships between benign models and their training data features. By manipulating these structural correlations, the attack maximizes the training loss of the federated learning system, compromising its overall effectiveness.

This paper focuses on Zero-Trust Foundation Models (ZTFMs), a novel paradigm that embeds zero-trust security principles into the lifecycle of foundation models (FMs) for Internet of Things (IoT) systems. By integrating core tenets, such as least privilege access, continuous verification, data confidentiality, and behavioral analytics into the design, training, and deployment of FMs, ZTFMs can enable secure, privacy-preserving AI across distributed, heterogeneous, and potentially adversarial IoT environments. We present the first structured synthesis of ZTFMs, identifying their potential to transform conventional trust-based IoT architectures into resilient, self-defending ecosystems. Moreover, we propose a comprehensive technical framework, incorporating federated learning (FL), blockchain-based identity management, micro-segmentation, and trusted execution environments (TEEs) to support decentralized, verifiable intelligence at the network edge. In addition, we investigate emerging security threats unique to ZTFM-enabled systems and evaluate countermeasures, such as anomaly detection, adversarial training, and secure aggregation. Through this analysis, we highlight key open research challenges in terms of scalability, secure orchestration, interpretable threat attribution, and dynamic trust calibration. This survey lays a foundational roadmap for secure, intelligent, and trustworthy IoT infrastructures powered by FMs.

Cloud-of-clouds storage systems are widely used in online applications, where user data are encrypted, encoded, and stored in multiple clouds. When some cloud nodes fail, the storage systems can reconstruct the lost data and store it in the substitute nodes. It is a challenge to reduce the latency of data recovery to ensure data reliability. In this paper, we adopt a Reinforcement Learning-based Data Recovery (RLDR) approach to reduce the regeneration time. By employing the Monte-Carlo method, our approach can construct the tree-topology-based regeneration process, a.k.a. regeneration tree, to effectively reduce the regeneration time. Through rigorous analysis, we apply the information flow graph to optimize the inter-cloud traffic for a given regeneration tree. To verify the merit of RLDR, We conduct extensive experiments on real-world traces. Experiments demonstrate that RLDR can significantly accelerate the regeneration process. Specifically, RLDR can reduce the regeneration time by up to 92% and increase the throughput by up to twelve-fold, compared to the prior art.

Edge computing, an advanced extension of cloud computing, provides superior computational capabilities and lowlatency processing at the network edge, facilitating its availability for real-time data analysis in resource-limited settings. When applied to the analysis of teaching methodologies, edge computing enables the seamless integration of vocal and physical cues, facilitating collaborative, dynamic, and real-time evaluations of teaching quality. However, the inherent complexity of human perception and multimodal interactions impose great challenges to the analysis of these aspects in Artificial Intelligence of Things (AIoT). This paper introduces an innovative mathematical model and a measurement index specifically designed to assess changes in voice-body coordination over time. To achieve this, we propose a cloud-enabled enhanced Bi-Linear Attention Network incorporating entropy and Fourier transforms (BAN-E-FT), which leverages both temporal and frequencydomain features. Specifically, by harnessing the computational and storage capabilities of edge computing, BAN-E-FT facilitates distributed training, expedites large-scale data processing, and enhances model scalability, where entropy measures and Fourier transforms capture modality dynamics, enhancing BAN's fusion capabilities. Moreover, a conditional domain adversarial network is embedded to address regional teaching variations, improving model generalizability. We also verify the robustness of BAN-EFT with accuracy and convergence through convex optimization analysis. Experiments on the eNTERFACE'05 dataset demonstrate 81% accuracy in assessing teaching adaptability, while real-world test at Guizhou University confirms 78% accuracy when using BAN-E-FT, matching human expert assessments.

The autonomous interpretation of application intent (APPI) represents the primary step towards achieving closed-loop autonomy in zero-touch networking (ZTN) and also a prerequisite for intent-based networking (IBN). However, understanding APPIs and invoking the corresponding network resources require network professionals with extensive technical expertise to customize network service requests (NSRs), which presents significant challenges for the large-scale deployment of ZTN. This paper investigates an interesting problem of autonomous interpretation of APPIs for ZTN, where a novel mechanism integrating hypergraph and transformer with completeness assurance (HyperTrans-CA) is proposed. In particular, we first involve the Bayesian theory to model APPIs interpretability as maximizing the correct transition probability, where hypergraph is used to describe the complex relationship between application characteristics (e.g., scenario function, and performance) and NSRs, including network devices, virtual network functions (VNFs), and resources. Then, the hypergraph is integrated into the encoder, decoder, and attention mechanisms of Transformer, and a completeness assurance mechanism is designed to improve the prediction accuracy. The convergence of HyperTrans-CA and the corresponding convergence speed of the hypergraph-boosted Transformer in the graph search process are also analyzed. Comprehensive simulations and empirical measurements regarding industrial internet demonstrate that HyperTrans-CA can effectively explain/understand APPIs. Compared to the state-of-the-art Transformer and ChatGPT3.5 models, HyperTrans-CA improves the prediction accuracy of APPIs mapped to VNFs by 23% and 46%, respectively, while raising the prediction accuracy of VNF locations by 8.6 and 17.3 times.

Cloud-of-clouds storage can enhance the data security and reliability of online applications by encrypting, encoding, and distributing user data across multiple clouds. Fast transferring large volumes of data through networks with limited bandwidths remains a practical challenge, especially in the event of disaster backup. To address this, we model a data storage process using an information flow graph and estimate inter-cloud traffic. We propose a new Network Coding-based Cloud-of-Clouds Backup (NC3B) framework, which enables collaborative encoding and data exchange among backup clouds to utilize inter-cloud bandwidth efficiently. We analytically corroborate that NC3B effectively reduces write operation latency. We also demonstrate the NC3B framework by incorporating two cutting-edge Reed-Solomon (RS) based data storage techniques, namely All-Or-Nothing Transform-RS (AONT-RS) and Converge AONT-RS (CAONT-RS), referred to as Network coding-based Backup AONT-RS (NBAONT-RS) and Network coding-based Backup CAONT-RS (NBCAONT-RS), respectively. To validate our approach, we deploy a real-world prototype storage system on Amazon EC2 using a cluster trace set, and underscore the effectiveness of NC3B, showcasing reductions in latency of up to 50% compared to state-of-the-art approaches, alongside throughput improvements of up to 98%. These findings underscore the benefits of NC3B in real-world storage scenarios.

This article investigates the issue of cooperative real-time trajectory design for multiple unmanned aerial vehicles (UAVs) to support emergency communication in disaster areas. To restore communication links rapidly between mobile users (MUs) and the ground base stations, UAVs equipped with both radio frequency (RF) modules and free space optics (FSO) modules are utilized as relay nodes. Given the challenges of setting up a central controller for the UAVs and the urgency of emergency communication, the trajectory design problem for these UAVs is formulated as a distributed cooperative optimization problem. Based on the enhanced${K}$-mean algorithm and multiagent PPO (MAPPO) algorithm, a cooperative trajectory design method, abbreviated as KMAPPO, is proposed for the UAVs to minimize interaction overhead and optimize deployment efficiency. Compared to the state-of-the-art deep reinforcement learning (DRL) methods, simulations reveal KMAPPO’s superior performance. It converges 32% faster, boosts RF allocation efficiency, and augments FSO communication backhaul capacity.

Targeted identification of essential proteins is of great significance for species identification, drug manufacturing, and disease treatment. It is a challenge to analyze the binding mechanism between essential proteins and improve the identification speed while ensuring the accuracy of the identification. This paper proposes a novel method called EPCRO for identifying essential proteins, which incorporates the chemical reaction optimization (CRO) algorithm and the naive Bayes model to effectively detect essential proteins. In EPCRO, the naive Bayes model is employed to analyze the homogeneity between proteins. In order to improve the identification rate and speed of essential proteins, the protein homogeneity rate is integrated into the CRO algorithm to balance between local and global searches. EPCRO is experimentally compared with 17 existing methods (including, DC, SC, IC, EC, LAC, NC, PeC, WDC, EPD-RW, RWHN, TEGS, CFMM, BSPM, AFSO-EP, CVIM, RWEP, and EPPSO-DC) based on biological datasets. The results show that EPCRO is superior to the above methods in identification accuracy and speed.

Abstract Domain Name System (DNS) is a key infrastructure on the Internet. The Distribution Master (DM) system is used to transmit root zone data from Internet Assigned Numbers Authority (IANA) to the root server. DM is a centralized system that will introduce single points of failure and abuse of authority. To solve the problem of a single point of failure, we adopt the decentralization of blockchain in architecture, and propose a blockchain-based distributed DM architecture, which allows nodes to join and exit at any time. At the technical level, the 3R-PBFT (Replicated, Redundant Practical Byzantine Fault Tolerance) algorithm is proposed to reach a consensus, which increases the security of the system. We use flexible mechanisms to reduce the number of signatures and improve the performance of the system. A threshold signature algorithm is used to ensure the uniqueness of the signature key information, and the increase of DM nodes will not bring about an increase in the number of keys. The advantages of the system structure in stability and scalability were verified by experiments.

Virtualized network function (VNF) orchestration dynamically deploys network slices, which provides an effective means of customized service provision. To achieve a realistic and comprehensive perspective of the decision process for customized service provision, we propose a virtualized resource orchestration strategy in the radio access network (RAN) of Internet of Things (IoT) based on chemical reaction optimization (CRO). Specifically, we apply particle swarm optimization (PSO), a Gaussian process, random walk model, and$Q$-learning to enhance the CRO algorithm to quickly obtain the approximate optimal solution for the proposed CRO-based resource orchestration strategy (CROROS). The simulation results show that compared with existing access methods, CROROS can reduce the service rejection rate of a virtualized RAN and improve the utilization rate of network system resources. Compared with other heuristic algorithms [e.g., PSO, genetic algorithm (GA), and CRO], CROROS can accelerate the global approximate optimal solution and improve the approximate fitness of the approximate optimal solution within a specified time.

With the development of informatization, network businesses are expanding and business functions are becoming more powerful; thus, network infrastructure has begun to face new challenges in serving businesses. Research regarding the dynamic deployment of network slices according to business requirements is urgently needed. To develop a realistic and complete view of service providers' decision-making process, adaptation costs must be considered. Function orchestration of virtualized radio access networks is a critical way for network slices to offer web apps customized service. In this paper, we develop a mathematically optimized model for function orchestration and propose an orchestration strategy using a form of particle swarm optimization (VNFPSO), in virtualized radio access network. Bearing in mind the openness of radio access network frameworks, discreteness of network functions, and proliferation of network traffic, we have enhanced the state-of-the-art PSO algorithm in terms of inertia weight, particle mutation, and factor learning to improve the speed of obtaining a global approximate optimal solution. Simulated experimental results show that this strategy could decrease rejection rates in virtualized radio access networks and improve the utility of network system resources.

Abstract The Internet of vehicles (IoV) is an emerging networking technology, which can support information sharing and interactions among users, vehicles, and infrastructures. Various applications can be provided by IoVs, and they have very different quality‐of‐service (QoS) requirements. It is a great challenge to design an efficient MAC protocol to meet the different QoS demands of various applications in IoVs, because of unreliable links and high vehicle mobility. On the other hand, cooperative communication is effective in mitigating wireless channel impairments by utilizing the broadcast nature of wireless channels. In this paper, a multi‐channel cooperative clustering‐based MAC (MCC‐MAC) protocol, under the Dedicated Short Range Communication (DSRC) multi‐channel architecture, is presented to improve the transmission reliability of safety messages and provision QoS for different applications in IoVs. Further, we analyze the performance of MCC‐MAC, in terms of average transmission delay. In addition, extensive simulations with ns‐2 are conducted to demonstrate the performance of the proposed MCC‐MAC. Copyright © 2016 John Wiley & Sons, Ltd.