Existing Signed Graph Neural Networks optimize link sign prediction objectives fundamentally misaligned with friend recommendation, while discarding trust asymmetry, edge strength, and adaptive social theory application. We present ASRA-GNN, addressing these gaps through three contributions: Sign-Aware Structural Role Attention (SSRA) grounded in four social network theories; a Locally Adaptive Theory Mixing (LATM) gate replacing TrustSGCN's binary global threshold with a continuous per-node end-toend learned mixing function; and a Signed Contrastive Recommendation Loss providing the first ranking objective for signed user-user graphs using observed positive-negative pairs as natural contrastive anchors. Experiments on Bitcoin-OTC and Bitcoin-Alpha demonstrate an average of Recall@10 of 0.0599, NDCG@10 of 0.4280, and Precision@10 of 0.0840, outperforming all other baselines.
The fast rise of wireless communication networks, including 6G, Internet of Things (IoT), and edge com puting, has created unprecedented demand for spectrum and energy resources.become a significant challenge in modern IoT networks due to heterogeneous devices, dynamic traffic patterns, and diverse QoS requirements. This study proposes a Deep Reinforcement Learning (DRL)–basedframework for optimizing network resource allocation in IoT environments using real-world sensor data. The proposed framework differs from existing studies that typically assess reinforcement learning methodologies under simplified wireless network assumptions and idealized conditions. Our method functions on heterogeneous IoT traffic produced by various device types, including sensors, actuators, and cameras, each possessing distinct Quality of Service (QoS) requirements. To ensure practical applicability, a realistic IoT simulation environment is developed, incorporating dynamic bandwidth release and queue-aware resource management to emulate real-world network behavior. Furthermore, a Deep Q-Network (DQN) agent with an enhanced exploration strategy is designed to improve learning stability and convergence performance, enabling more efficient and adaptive resource allocation in dynamic IoT scenarios. Experimental results show that the proposed DQN agent achieves a 26.7% improvement in cumulative reward compared to a random policy and consistently outperforms conventional heuristic approaches. This significant gain indicates that the agent effectively learns a structured resource allocation strategy rather than making uninformed decisions. These results confirm that reinforcement learning–based resource allocation provides a scalable and effective solution for IoT networks, particularly in environments characterized by large state spaces, dynamic network conditions, and stochastic traffic patterns.
The rapid proliferation of Internet of Things (IoT) devices has placed unprecedented pressure on the network edge, where applications such as augmented reality, real-time analytics, and autonomous navigation demand low latency and tight energy budgets that traditional cloud-centric architectures cannot meet. Multi-access Edge Computing (MEC) addresses this gap by relocating computation closer to end users, but the core question of where and how each task should be executed remains open: rulebased and single-objective offloading strategies fail to simultaneously balance service latency, energy efficiency, and user experience under dynamic, large-scale conditions. In this paper we propose TARLOT (Two-Agent Reinforcement Learning Offloading Tasks), a cooperative framework for threetier IoT–MEC–Cloud environments. TARLOT decouples the offloading decision from the resourceallocation problem and assigns each to a dedicated Q-learning agent, so that the two subproblems are specialised independently while still being optimised jointly. The framework is evaluated on PureEdgeSim under heterogeneous IoT workloads, device densities ranging from 200 to 2,400, and diverse application profiles, and is compared against five widely-used baselines (Random, Round-Robin, Trade-Off, Pure-Edge, and Pure-Cloud). At 2,400 devices, TARLOT delivers an average service time of 1.1 s (against 4.3 s for Pure-Cloud), a Quality of Experience of 0.77 (against 0.22 for Pure-Cloud), a task-failure rate below 2 % (against nearly 14 % for Pure-Cloud), and a per-device energy consumption of only 3.6 W (against 11.2 W for Pure-Cloud) — roughly a 68 % reduction. Balanced CPU utilisation across the local, edge, and cloud tiers further confirms that TARLOT prevents resource bottlenecks, establishing it as a practical solution for next-generation large-scale IoT deployments.
Multi-Access Edge Computing (MEC) brings computation closer to end users to reduce latency and energy consumption for compute-intensive mobile applications. In this paper, we address the joint task offloading and resource allocation problem in multi-user MEC systems and propose a decentralized control framework based on Multi-Agent Reinforcement Learning (MARL). Each user–base station link is modeled as an autonomous agent that decides whether to execute tasks locally or offload them to the edge, and how much computing resource to request. The learning framework follows a centralized training and decentralized execution paradigm and integrates Q-Learning, Deep Q-Network (DQN), and Double DQN (DDQN) algorithms. Extensive simulations show that the DDQN-based approach achieves lower total system cost and faster convergence than the full-local, full-offload, and heuristic baselines. The results confirm that lightweight MARL is a practical and scalable solution for dynamic MEC environments under realistic resource constraints.
Although distributed systems can exist across multiple data centers, they require fog and cloud computing paradigms for data management in an Internet of Things (IoT) era. The integrated IoT, fog, and cloud (IoT-fog-cloud) method enables the processing of large amounts of IoT data in real-time. Alternatively, a lack of Load Balancing (LB) and improper handling of network resources can reduce the Quality of Service (QoS) under such circumstances. In real-time applications, increasing traffic to fog nodes causes delays and increases energy consumption. This problem was resolved by an effective LB algorithm. However, good resource utilization can be achieved whenever an effective LB is incorporated with Task Scheduling (TS). Hence, this article proposes a Multi-Objective Weight Optimized Task Scheduling and Load Balancing (MOWOTSLB) for IoT-fog-cloud systems. This research aims to effectively schedule workloads in a balanced manner, which conserves energy, enhances QoS, and reduces task execution time. This scheme employs a Horse Herd Optimization Algorithm (HOA) for TS. This HOA optimizes the scheduling of users’ task requests to suitable computing resources according to the fitness function calculated using makespan, execution cost, and energy utilization. Though it improves resource utilization, some Physical Machines (PMs) are overloaded, and others are underloaded during uncertain fluctuations in workloads. This causes high energy and resource wastage in the data center. To solve this problem, the HOA is also adopted for Virtual Machine (VM) migration in this article. By assessing the fitness function of PMs such as load, migration cost, energy, and bandwidth use, the HOA chooses the finest VMs to drift to the appropriate PMs. This successfully strikes a balance between load distribution among PMs and energy usage. The simulation findings show that the MOWOTSLB outperforms current LBTS schemes in 0.95 throughput, 30 ms delay,100 ms response time, 175 kj energy consumption, 12.5 mb memory usage, and 900s lifetime.
The modern cloud computing systems have to plan the heterogeneous workloads and balance performance effectiveness, service availability, and sustainability. In this study, an adaptive hybrid scheduling framework is developed Adaptive Ant-guided Min-Max (AAMM) combining ant-guided optimization with dynamic Min-Min and Max-Min in deciding how to allocate cloud tasks as a multi-objective. The scheduler jointly evaluates task completion time, the likelihood of Service Level Agreement violations, energy consumption, and monetary cost within a unified scoring framework, enabling informed trade-offs among competing objectives. AAMM is assessed based on a real disaggregated Deep Learning Recommendation Model workload of 1,544 heterogeneous tasks, running on heterogeneous virtual machines. Comparative experiments are done with Min-Min, Max-Min and ACO-guided Min-Min scheduling strategies. According to experimental findings, the suggested approach has been very effective in reducing energy per task, cost per task, SLA violations are significantly lowered, and flow time stability is enhanced. Though moderate growth in the makespan is witnessed, the accompanying trade-off has created equal distribution of resources and service reliability.
Wireless Sensor Networks (WSNs) are being used more for monitoring and surveillance, where nodes with limited energy resources must send a lot of sensed data. One big problem with these kinds of networks is that nodes run out of power quickly because they send the same data repeatedly. This has a direct impact on the network's lifetime and reliability. Many data aggregation methods have been suggested to cut down on communication overhead, but most of them only deal with redundancy between nearby nodes and does not do a good job of stopping repeated transmissions from the same node over multiple sensing rounds. Also, security concerns are often not fully considered in redundancy elimination systems. This paper introduces a Robust and Efficient Redundancy Elimination Secure Data Aggregation (RERESDA) model for clustered Wireless Sensor Networks (WSNs) to address these limitations. The suggested method presents a pattern-based data representation system that takes advantage of changes in sensed data over time. Sensor nodes only send data to the cluster head when they notice a change in the pattern they are making. This keeps them from sending data that is not needed. Also, when a cluster has similar data patterns, the cluster head picks a representative node based on how much energy is left in it. This keeps data safe while making sure that energy use is balanced. We use MATLAB-based simulations to test how well the proposed scheme works with a network of 40 sensor nodes spread out over a 100 m × 100 m area. Experimental results indicate that the proposed model diminishes overall energy consumption by as much as 56% in comparison to non-aggregation methods, while concurrently reducing bandwidth utilization. The results show that RERESDA really does improve energy efficiency and network lifetime by getting rid of redundancy and safely aggregating data at the same time.
Cloud computing has revolutionized modern-day digital infrastructure for large-scale application deployments. However its widespread adoption has brought in significant cybersecurity vulnerabilities also which also pose new and emerging threats to native cloud environments.Although there are many attack detection models, their suitability and stability are weakening day by day due to the lack of cloud-native or newer datasets which are designed to contain newer attack data. Our cloud native attack dataset sets a new standard towards a dataset including 22 of the most common attacks faced by cloud servers and instances. A visual analysis of those attacks is generated to provide a visual difference between the behavior of the various attack types by classifying them into clusters based on the nature of attacks and through usage data of various parameters such as port usage and protocol usage. In this paper we have designed a cloud-native dataset creation model generated with controlled cloud instances to adopt a cloudnative approach and the visualization of the generated attacks. Hence, this cloud-native attack dataset can surely be helpful to the research community in validating and training their new models
The quick spread of Internet connections has instigated the revolutionary age of Cyber-Physical Systems (CPS) and Internet of Everything (IoE) devices. The IOE and CPS devices are the cornerstone of Industry 4.0. which is centred on Machine-to-Machine (M2M) communication. IoE and CPS devices are used in hostile environments and have limited computing and energy resources. Criticality and dependence of the Internet have exposed IoE and CPS systems to cyber-attacks. Thus, to prevent any damage, these systems require a competent and lightweight intrusion detection system (IDS). The current research recommends a novel IDS built upon a new feature selection algorithm which can identify entropy reducing and highly statistical reliable features from a dataset. The proposed feature selection technique showed significant improvements in performance measures for several classifiers. Proposed IDS with the IOTID20 dataset demonstrated that the accuracy and performance metrics exceeded 99%. The trustworthiness of the proposed IDS is further supported by its constant efficacy on the NSLKDD dataset. The proposed IDS is found to be competitive with all previous studies in all performance areas. Thus, proposed IDS on novel and innovative feature selection techniques can protect the digital ecosystem and IoE landscapes from cyber-attacks to bolster Industry 4.0.
The proliferation of 6G networks poses new challenges to traditional methods of network management due to the massive volumes of data generated and the wide variety of devices they link. A paradigm change towards AI frameworks built in Machine Learning (ML) and Deep Learning (DL) is essential due to the shortcomings of these approaches. A Speed-optimized Attention-based Hybrid Graph Convolutional Network-Long Short-Term Memory (SPAH-GCN-LSTM) model and a Reinforcement Learning (RL) framework utilizing Q-Learning (QL) were developed to forecast network congestion and enhance data transmission routes, respectively. Nevertheless, in a dynamic network, uncertainty in routing decisions could be caused by the switching between policies by a single agent. The time spent training one agent is prohibitive with the increase of the size of the network. In spite of the fact that multi-agent RLhas been utilized to alleviate this problem, classical QL can potentially face the challenge of non-stationarity that arises because of the joint learning of other agents in a multi-agent stochastic-game environment. As a result, the present manuscript presents a Multi-Agent Multi-Step Deep QL(MAMS-DQL) system that is aimed at optimizing Washington routes in 6G networks. The main goal is to come up with a decentralized mechanism where every agent is able to choose its best routing strategy independently. The multi-agent dueling deep Q-network architecture is followed in this method so as to optimize routing decisions and identify the most efficient route of the network. It also uses a multi-step experience-replay strategy, which allows agents to modify their routing strategy by taking advantage of multi-step experiences across consecutive time steps of training. Lastly, the outcomes of the simulator show that the MAMS-DQL has a higher routing efficiency compared to traditional reinforcement-learning approaches.
Routing is an important element of network communication that enables data transmission between network devices. This process depends on a routing protocol, which assesses and selects the best route for data communication from source to destination using routing metrics. Consequently, the network's performance is directly impacted by the routing metric selection. Therefore, the choice of a routing metric directly influences the performance of the network. For example, an inadequate routing metric can increase data packet loss, latency, and energy consumption. In this study, we recognise the importance of routing metrics. For instance, the Expected Transmission Count (ETX) is a good measure of link quality. It estimates the transmission count required for successful communication. However, even though ETX is effective in terms of link reliability measurement, it fails to consider the consumed energy during communication. Yet, every data transmission consumes energy because a path with high reliability can still consume a lot of energy due to its length. For this reason, we propose Energy Aware_ ETX (EA_ETX), an enhanced form of ETX, which considers both link quality and energy consumption during communication. We integrate this proposed routing metric in the Routing Protocol for Low-Power and Lossy Networks (RPL) and propose an objective function, which is labelled OF_EA_ETX. We conducted simulations in Contiki Cooja and compared the results of OF_EA_ETX with the predefined objective functions of RPL (MRHOF and OF0). The results demonstrate that our approach outperforms OF0 and MRHOF, respectively, by reducing energy consumption by 11.29% to 28.48%, minimising end-to-end delay by 2.08% to 21.58%, and improving packet delivery ratio by 16.67% to 36.13%.
Obfuscation has been increasingly difficult in the subject of cybersecurity, since malware developers use it to change code appearance without changing its malicious behavior. As a result, signature-based and basic heuristic detection systems are easily bypassed by these techniques. This article reviews recent and ongoing research in the analysis and detection of obfuscated malware, giving special attention to methods that were recently developed to address this problem. The reviewed methods are divided into five major classes: static analysis, dynamic analysis, hybrid analysis, machine learning, and deep learning. thirty-six recent research papers from 2018 to 2025 are analyzed, with a detailed summary of each, including merits and demerits. The review is intended to generate a broad picture of the research field, point out strengths and weaknesses in each category, and identify the way forward, especially for the area of hybrid and deep learning-oriented memory analysis.
In this age of the Internet of Things (IoT), crucial management frameworks for large, dispersed systems must include fog and cloud computing technologies. By combining IoT with fog and cloud computing, massive amounts of IoT data can be processed in real time, meeting all of your processing needs. Nonetheless, optimizing Resource Allocation (RA) and Load Balancing (LB) in dynamic and varying operating settings continues to be a critical issue. Many traditional RA-LB techniques in IoT-fog-cloud systems frequently experience dynamic and unpredictable workload variations, leading to the overloading of some Physical Machines (PMs) and the underutilization of others. This disparity might result in high energy usage and resource scarcity in fog-cloud data centres. This study presents a novel Optimized Virtual Machine (VM) Migration-based RA-LB (OVM2-RALB) technique utilizing the Enhanced Starfish Optimization Algorithm (ESOA) for IoT-fog-cloud systems. To drastically cut down on energy consumption in fog-cloud data centers, the main objective is to dynamically move VMs from overcrowded PMs to underused ones. At first, the incoming job is categorized as eitherfog-dependent or cloud-dependent based on its guaranteed ratio. This ratio is decided by available PMs and their corresponding VMs in fog and cloud. The mean load for each PM is computed, which is utilized to calculate the load balancing factor to identify overloaded and underloaded PMs. Then, the ESOA, which is an improvement upon the standard SOA by combining tent chaotic mapping and Logarithmic Spiral Reverse (LSR) learning, is adopted for the VM migration process. This ESOA seeks to choose the most suitable PM for VM migration. It also determines the most appropriate VM for migration based on migration expenses, load balancing factor, energy use, and bandwidth utilization. Furthermore, the selected VM in the overloaded PM is migrated to the chosen underloaded PM. Thus, this VM migration results in a balanced load and alleviates overload on the PM. Finally, simulation results show that this OVM2-RALB outperforms conventional RA-LB techniques in IoT-fog-cloud environments.
This paper proposes a distributed cloud middleware system that offers Offloading as a Service (OaaS) for mobile applications, aiming to satisfy user Quality of Service (QoS) requirements while reducing response time and overall cost. OaaS takes into consideration different features such as the user location (proximity to cloud resources); application requirements (e.g., OS, RAM, number of vCPUs); and desired QoS to offer an efficient offloading service that maximizes cloud resource utilization and reduces Virtual Machine (VM) rental costs. To ensure cost efficiency, OaaS dynamically selects VMs from various cloud providers based on pricing and adjusts the number of VMs in real-time to meet response time requirements. Using a predictive model based on queuing theory, the middleware can scale the VM pool up or down by forecasting workload demands. Simulation results confirm that our model outperforms existing algorithms in terms of response time and VM leasing costs, while meeting users’ QoS requirements.
Enabling healthcare services over emerging Sixth Generation (6G) networks and Internet of Things (IoT) introducesa strict requirementthetimely and reliable allocation of medical resources. Prediction of resource allocation efficiency based on rule-based or manual policies often fails to be adaptive to heterogeneous demands and dynamic conditions of IoT networks. To address this challenge, a multi-model regression-based approach is proposed to predict the efficiency of resource allocation for optimizing the MR infrastructures of IoT and 6G networks. The approach consists of data pre-processing, exploratory data analysis, multi-model regression learning, and operational factors interpretation. First, the dataset is loaded and non-informative identifier attributes are removed to reduce noise and improve generalization. Correlation analysis is performed through a heat map plot of numerical features to identify features that are strongly related to the target variable. Extensive experiments are conducted on a publicly available dataset to evaluate the proposed approach according to a number of performance metrics, such as the root mean square error (RMSE), determination coefficient (R-squared), and mean absolute error (MAE). Experimental results showed that the best regression model of proposed approach attains the highest prediction performance compared with other models and state-of-the-art work. In addition to predictive superiority, interpretation of best model’s outputs regarding to throughput and utilization of the network is reported to show the association between predicted efficiency, network speed, and utilization status, which will help to design an actionable plan for deploying intelligent allocation policies.
Advanced adversaries use unique initial access strategies establish persistence in corporate systems. This research presents behavior-driven threat hunts focused on phishing via malicious Microsoft Word documents, using living-off-the-land binaries (LOLBINs) and subtle alterations to Remote Desktop Protocol (RDP) services for stealthy lateral movement. Elasticsearch SIEM was used to ingest and analyze 303,148 logs utilizing Kibana and Lucene-based detection queries. The initial investigation revealed 44 connected instances in which Winword.exe initiated cmd.exe, resulting in the download of a dubious payload (MicrosoftUpdate.exe). Persistence was validated via Sysmon Event Code 11 (file creation in the Windows Startup directory), whereas lateral movement and command-and-control operations were indicated by outbound connections on port 9000. The second hunt identified 20 registry-related events that verified the alteration of the RDP port from the default 3389 to 3398 via reg.exe, succeeded by remote interactive logon (Event ID 4624, Logon Type 10). Subsequent network activity indicated a connection via FTP port 21, suggesting possible data exfiltration. To augment scientific rigor, the authors incorporated mathematical validation through Bayesian inference, Threat Confidence Scoring, and entropy-based behavioral diversity analysis. The phishing scenario attained a posterior threat probability of 0.820, whereas RDP port manipulation resulted in 0.778. The calculated TCS attained 20, signifying a severe multi-stage compromise. Entropy analysis revealed 2.181 bits (~84% of maximal entropy), indicating substantial event variety aligned with coordinated adversarial actions. This research's primary contribution is the integration of behavioral SIEM query logic with probabilistic validation and entropybased complexity modeling, enabling SOC teams to prioritize alerts based on quantitative threat confidence rather than relying solely on signature-based detection.
Cybersecurity language models are typically evaluated under fragmented ways, which impedes meaningful comparison and operational comprehension. Existing cybersecurity-specific BERT models are often tested in isolation, with inconsistent preprocessing, tokenization, and evaluation procedures. This paper presents a unified cross-domain benchmarking study for systematically evaluating cybersecurity-adapted BERT models under identical experimental conditions. The evaluation spans CTI, phishing, logs, and CVE domains using CTI-BERT, SecureBERT, CySecBERT, and SecBERT. Results reveal strong performance convergence across models and highlight domain-driven failure modes rather than architectural superiority. To examine real-world resilience, the study expands on this paradigm with zero-shot and fewshot cross-domain evaluations, revealing asymmetric transfer behavior and domain-dependent adaptation efficiency. A controlled training method ablation is also performed, indicating that aggressive optimization does not always increase performance and can decrease stability in semantically rich domains. Stress filtering further exposes brittle reliance on lexical shortcuts and limited semantic grounding. These findings provide practical guidance for model–domain alignment and real-world cybersecurity deployment. The findings of this study are especially relevant to network security and operational situations when cybersecurity models are deployed across heterogeneous data streams
This article presents an Edge-AI enabled hybrid deep learning framework for real-time fault detection and root cause analysis in industrial motors using multimodal sensor data, containing vibration, temperature, and current signals. The proposed system leverages a CNN-LSTM model deployed on edge devices to accurately classify operational states into normal, minor, and major faults. A dataset comprising 15,000 labeled samples collected from real-world industrial setups and augmented through techniques such as SMOTE and time-warping was used for training and evaluation. In the implementation,the CNN component captures spatial patterns in sensor data, while the LSTM layer models temporal dependencies, enabling effective fault diagnosis. The proposed hybrid model achieved superior performance with 96.8% accuracy, 97.2% precision, 96.5% recall, and an F1-score of 96.8%, along with a low inference latency of ≤198 ms, demonstrating suitability for real-time edge deployment. Comparative analysis against CNN-only and LSTM-only models confirms the hybrid architecture’s advantage in fault sensitivity and prediction reliability. Additional insights from confusion matrix analysis, ROC-AUC evaluation, and fault-wise performance metrics validate the model’s robustness. The system also incorporates TinyML-based optimizations and lightweight messaging for efficient edge computing, making it a scalable solution for predictive maintenance in Industry 4.0 applications.
Wireless sensor networks (WSNs) and Internet of Things (IoT) networks have applications in smart cities, industrial monitoring, healthcare, and environmental surveillance, but since the networks have open wireless communication, resource constraints, and decentralized architecture, they are very susceptible to security threats such as malicious routing, manipulation of trust, and interference with data. Although cooperative multipath routing enhances fault tolerance and reliability via distributing traffic among more than one path, identifying correct evaluation of trust and ensuring the safety of coordination between the involved nodes are critical factors of success. Currently used trust based routing methods are generally centralized or a model of trust that is not affected by dynamic attacks, which restricts the ability to scale as well as the ability to withstand such attacks. The present paper suggests a decentralized blockchain-based trust and security system in cooperative multipath routing within the WSN and IoT systems. The framework combines lightweight blockchain technology and cooperative routing in order to offer safe, transparent, and tamper proof trust management. Network nodes keep an unchangeable distributed registry of routing behavior, updates of trust, and history of cooperation, which means that malicious nodes cannot lie or modify trust information. Smart contracts apply to compute the trust and route validation depending on the key performance indicators, including packet forwarding success, compliance with latency, and consistency of behavior, which are useful in dynamically determining reliable routing paths. Results obtained through simulation shows that the suggested framework has a better packet delivery ratio, routing stability, and resistance to blackhole, gray hole, and Sybil attacks than regular trust-based routing schemes, and has acceptable computational and communication overhead. This is because these findings demonstrate that the solution under consideration provides a scalable, cost-effective, and secure means of cooperative routing in next-generation WSN and IoT networks.
During the last decade, there has been a massive development of wireless networks, and nowadays 4G and 5G technologies are a usual thing. The next generation 6G standard is even more promiscuous, such as improved artificial intelligence (AI). In order to maximize resource allocation on 6G networks, the study suggests Horned Lizard Ensemble VotingResource Allocation (HLEVRA) model. HLEVRA uses AI methods to compute user needs on resources and distribute them to them. In simulating an environment of a 6G network using NS3, the HLEVRA performance is measured according to the most important parameters, including throughput, data transfer rate, energy consumption, communication delay and packet drop. The findings reveal that HLEVRA is effective in the management of resources in a 6G network. HLEVRA recorded a spectacular throughput of 14.7 Gbps, data rate of 840 Kbps, 0.33 mW of energy usage, 20 ms of communication delay and packet drop rate of 12.5 with 100 users connected. These results indicate that HLEVRA is an effective strategy to optimize the 6G network performance.
