Data collection is the most critical capability of wireless sensor networks. Recent studies have shown that sink mobility along a fixed path can improve energy efficiency and simultaneously lead to high network throughput. However, achieving these improvements results in increased data transmission delay. To address this issue, we propose a novel data collection scheme in which all sensed data must be collected within a given time constraint while ensuring that the energy overhead is minimal. This novel data collection model is first formulated as an integer linear programming problem and then solved using an effective hybrid search method named probabilistic learning-based optimal node assignment (PLONA). The method integrates a reinforcement probability learning-based framework to guide the search toward new promising regions and a mixed feasible and infeasible search coupled with a diversified late acceptance strategy. Additional features of PLONA include an innovative transplant for generating diversified, high-quality initial solution and a streamlining technique to accelerate candidate solution evalution. The proposed PLONA algorithm is rigorously evaluated on extensive public benchmarks. The results demonstrate that PLONA matches CPLEX on all 22 small and medium-scale instances, outperforms CPLEX on 15 of 22 large-scale “DataColl” instances and on the majority of the 36 independent benchmark cases, and surpasses the state-of-the-art heuristic MDGOSP across most configurations while requiring substantially less computation time.
Intent-Based Networking (IBN) is a key paradigm in sixth-generation networks to manage the heterogeneity, scalability, and dynamism of resources across the Cloud Continuum. In distributed environments, resource allocation and orchestration are hindered by the trust among end-users and the complex chains of services between multiple stakeholders and domains. Traditional Service Level Agreements (SLAs) are insufficient to address these dynamics, requiring adaptations to incorporate trust characteristics (Trust Level Agreements, TLAs) that define trustworthiness thresholds, monitor real-time variations, and enforce compliance throughout a relationship. Accordingly, this article presents a trust-aware monitoring framework that establishes a monitoring-to-Knowledge Graph (KG) pipeline to track intent compliance. The framework leverages telemetry from Cloud Continuum resources, following RFC 9417 for service assurance, to evaluate operational states and assign health scores that lay the foundation for future trustworthy service selection. A tailored trust-oriented ontology and KG ensure semantic alignment, thereby enhancing advanced reasoning capabilities and a broader pattern recognition. The framework is validated under attack-scenario simulation experiments, achieving promising results (0.771 ± 0.043 F1-Score, 0.771 ± 0.090 precision, and 0.810 ± 0.041 recall) in detecting and evaluating relevant changes in service health. These results show its feasibility and potential to enhance trust-aware automation and intent compliance in 6G IBN environments.
The convergence of Artificial Intelligence, the Industrial Internet of Things (IIoT), and collaborative robotics in Industry 5.0 exponentially multiplies attack surfaces while dissolving the well-defined perimeters on which traditional security models rely. Zero Trust Architecture (ZTA), grounded in the principle of never trust, always verify, reorients industrial cybersecurity from boundary defence to continuous identity verification, dynamic access control, and least-privilege policy enforcement across every device, user, and data flow. Despite growing adoption interest, ZTA research for IIoT remains fragmented across isolated mechanism categories and lacks a unified synthesis spanning the full spectrum of Industry 5.0 constraints, including brownfield legacy integration, real-time determinism, federated learning, and quantum-resilient cryptography. This systematic literature review, conducted according to PRISMA 2020 guidelines, synthesises 45 peer-reviewed original articles selected from an initial pool of 492 documents published between 2022 and 2026. A five-category ZTA taxonomy is developed covering authentication and access control, anomaly detection and intrusion prevention, blockchain-enabled trust management, federated and split learning, and post-quantum and hardware-enforced security. A 15-dimension coverage analysis reveals persistent gaps in legacy-compatible adaptive trust models and blockchain-secured distributed data storage. Quantitative synthesis finds that ZTA implementations achieve mean authentication latency of 35 ms, IDS accuracy of 95.2% peaking at 99.9%, and blockchain-enabled federated learning consensus within 250 ms, while gateway-optimised architectures demonstrate communication latency reductions of up to 73%. These findings point toward IEC 62443-compliant, gateway-based adaptive ZTA proxies integrated with blockchain-secured distributed storage as the priority research direction for brownfield Industry 5.0 deployments.
The sudden increase in the number of IoT and IIoT devices in software-defined networking (SDN) environments has been shown to be highly vulnerable to Distributed Denial-of-Service (DDoS) attacks, which can affect the operation of controllers and reduce the network's performance. Current protective measures, especially those employing two-level threshold systems, fail to address the issues posed by diverse and constantly changing IoT and IIoT systems. Therefore, in this research, we introduced a novel Adaptive Three-Tier Defense (A3TD) framework. It combines multi-level threshold protection, advanced feature extraction, intelligent detection, and adaptive mitigation. Traffic flows are preprocessed and refined for analysis. Features are extracted with a Dual Path Pyramid Vision Transformer (DP-PViT) approach. The A3TD module employs a three-level threshold system, enabling earlier detection of abnormal activities and providing fine-grained observation. It utilizes an Optimized Fuzzy Deep Reinforcement Learning (FDRL) model to classify streams in each detection phase. The Improved Shark Smell Optimization (ISSO) algorithm optimizes hyperparameters, resulting in efficient convergence and enhanced detection. The mitigation module utilizes the Safe-to-Block List Ratio (SBL), Flow-Table Occupancy (FTO), Packet-In Rate (PIR), and Flow-Rule Installation Rate (FRIR) to identify malicious and lawful flows. It applies graduated counteraction based on the A3TD thresholds. The framework has been examined on five recognized datasets: CICIoT2023, InSDN, IoT-23, RT-IoT2022, and WUSTL-IIoT2021, and demonstrates superior accuracy, precision, recall, latency, throughput, jitter, and F1 score. The system combines vision transformers for feature extraction, optimized reinforcement learning for detection, and adaptive mitigation to provide real-time, robust, and resource-efficient defense against advanced, persistent, and evolving attacks for SDN-enabled IoT and IIoT networks.
