Despite the recent increase in aquaculture production, information and communication technologies can significantly improve efficiency and help achieve better results. This systematic literature review studies previous research efforts on the application of intelligent environments to improve freshwater shrimp farming. The reviewed research works’ main contributions and opportunities for improvement are critically analysed to highlight the main benefits and existing gaps in using intelligent environment solutions for freshwater shrimp farming. The review highlights the importance of developing customised systems that address the specific characteristics of different freshwater shrimp farming processes, which are strongly influenced by the environmental context in which they occur, the species used for Pond grow-out, the users’ IT literacy, and other indicators that condition the technology to be used. This personalisation also affects the use of analytics techniques, which must adapt and perform with limited available data representing specific requirements of the process they support. The article proposes an architecture to guide the development of intelligent environments for freshwater shrimp farming, addressing the identified gaps and focusing on customising the behaviour of the system controlling the farming process.
Data centers (DCs) have become a fundamental infrastructure in people’s daily lives, providing remote data services for business, entertainment, and many other human needs, such as healthcare, smart cities, and smart homes. These large data centers must meet several dependability requirements to ensure quality of service with high reliability and availability. Data centers have three main infrastructures: IT (Information Technology), electrical, and cooling. Therefore, for data centers to achieve high availability, these infrastructures must be designed according to redundancy specifications ranging from Tier I to Tier IV. In this paper, we adopt a hierarchical and heterogeneous modeling approach that adopts reliability block diagrams (RBDs) with stochastic petri nets (SPNs) to represent these redundant data center infrastructures. We performed an evaluation focusing on key metrics, including availability, reliability, downtime, and uptime, as well as perform a sensitivity analysis of the constructed models to verify which subsystem most impacts the system’s behavior. Finally, we observe that the values obtained for the evaluated metrics using the proposed models fall within the reference parameters, indicating that real-world scenarios were successfully represented in the tier analysis. Furthermore, the sensitivity analysis showed that the subsystem most impacting system availability is the cooling subsystem.
Drone-based delivery systems have emerged as a promising solution for urban logistics, offering fast, flexible last-mile delivery. However, fluctuations in demand for requests over time make fleet sizing challenging. Overprovisioning leads to resource underutilization and higher operational costs, while underprovisioning may lead to delivery delays and violations of Service Level Agreements (SLAs). To address this challenge, this work proposes Drones-DT, a Digital Twin architecture designed for dynamic drone fleet management. The approach relies on a Stochastic Petri Net (SPN) analytical model synchronized with a delivery simulator to maintain a continuously updated Digital Shadow of the system. Based on real-time metrics, the Digital Twin performs predictive what-if analyses to estimate performance indicators, such as Mean Mission Time (MMT), system throughput, fleet utilization, and energy consumption, across different fleet configurations. A SLA-oriented decision mechanism, implemented through a binary search strategy, determines the minimum number of drones required to satisfy the target delivery time. The SPN model was statistically validated against a discrete-event simulator, yielding results equivalent to those of the simulator for MMT and throughput at the 95
In IoT intrusion detection, neural classifiers often produce overconfident predictions on novel or rare threats, undermining reliability and operational trust. This challenge highlights the urgent need for models that not only detect intrusions accurately but also express calibrated uncertainty to flag ambiguous or unfamiliar patterns. We propose GIBBON, a Graph-Based Intrusion Bayesian BNN with Gibbs, that integrates Graph Neural Networks (GraphSAGE) with Bayesian inference via Monte Carlo (MC) dropout and a blocked Gibbs sampling refinement for principled uncertainty estimation. GIBBON constructs relational graphs from raw network flow data to capture structural dependencies between hosts and connections. It applies MC dropout for variational posterior approximation and further refines the posterior using a Metropolis-within-Gibbs sampling scheme that iteratively samples classifier layer weights to improve uncertainty calibration. Evaluated on the complex NF-ToN-IoT-v3 dataset, our model achieves high predictive performance (accuracy = 0.8896, F1-score = 0.8845) while significantly outperforming prior IDS models in uncertainty calibration (ECE = 0.0055, MCE = 0.0217) and out-of-distribution detection (AUROC = 0.8832, AUPR = 0.4056, FPR@95
Remote sensor deployments often suffer from limited maintenance and connectivity issues due to adverse environmental conditions. Consequently, sensors are frequently in short supply, resulting in critical gaps in environmental data collection. Low-Power Wide-Area Network (LPWAN) protocols address these challenges by enabling long-range, low-energy communication. In this context, LoRaMesh networks rely on autonomous router nodes to maintain multi-hop connectivity in remote environments. Therefore, the dependability of individual nodes becomes a critical factor for sustaining reliable network operation. To this end, we develop a formal modeling approach using stochastic Petri nets (SPN) to evaluate the dependability and energy-driven availability of an autonomous LoRaMesh router node equipped with multiple energy sources. The results indicate that the modeled node maintains steady-state availability above 98
Unmanned aerial vehicles (UAVs), commonly known as drones, have played a crucial role in applications ranging from environmental monitoring to rescue operations. In recent years, their use in commercial services, especially for deliveries, has become an area of great interest due to the potential to reduce costs, increase transport speed, and expand reach in hard-to-reach locations, including densely populated urban areas. This advance represents a significant development in the evolution of logistics, introducing new operational paradigms that connect technology and efficiency. Yet, the use of drones in urban environments is not without its challenges, particularly in the realm of safety. The high density of obstacles, the presence of multiple aerial devices, and the unpredictable nature of environmental conditions all contribute to an increased risk of collisions. This risk, if left unaddressed, could significantly impact the performance and reliability of operations. It is therefore imperative that we investigate existing approaches and propose strategies that seamlessly integrate technological advances with operational safety. This study presents a systematic literature review to explore the causes of collisions and prevention strategies in the context of drone delivery services. From an initial analysis of 533 papers, 34 primary studies were selected for a detailed assessment. The work classifies the adopted approaches, such as types of delivery, collision types and causes, forms of detour, as well as the papers contributions and adopted evaluation. Based on the presented review, we highlight the research challenges and remaining research gaps. The analysis results in a comprehensive taxonomy and visualizations that help in understanding the state of the art, pointing out future directions to improve the safety and efficiency of drone delivery operations.
Augmented reality (AR) applications are revolutionizing healthcare, engineering, and education sectors, enabling users to integrate virtual elements into real-world environments seamlessly. Despite advancements in mobile computing, AR systems face significant challenges due to the computational limitations of mobile devices, particularly in scenarios demanding low latency and high reliability. Offloading computationally intensive tasks to edge servers presents a promising solution, but also introduces complexities in ensuring dependability. This paper proposes a process for evaluating the availability and reliability of AR systems with edge offloading. We analyze various redundancy strategies and their impact on system dependability by leveraging hierarchical modeling techniques, including reliability block diagrams and continuous-time Markov chains. Our findings reveal that adopting hot-standby redundancy and optimizing critical components, such as network infrastructure and mobile device batteries, significantly enhances system availability. The study concludes with a detailed case analysis, demonstrating how the proposed models can guide infrastructure planning for AR applications. This work provides actionable insights for developers and researchers seeking to design robust AR systems that can meet the stringent dependability requirements of real-world deployments.
Agriculture is a foundational industry that supports global food security, economic stability and environmental health. It encompasses the cultivation of crops and the raising of livestock, playing a critical role in providing essential resources and services. As the world faces increasing challenges such as climate change, resource depletion and population growth, there is a pressing need for innovative solutions to enhance agricultural productivity and sustainability. Our advanced agricultural monitoring system, the Sustainable Agri-tech monitoring system (SAMS), combines Internet of Things (IoT) devices and Tiny Machine Learning (TinyML) technology to tackle critical issues in the farming industry. SAMS automates water management, optimizes irrigation and monitors environmental conditions, all powered by a sustainable energy system. The system is designed for future integration with drone technology to enhance remote monitoring, early disease detection and overall resource efficiency. We also developed a novel TinyML-based model, DeepNet, which achieved 97.53
This paper presents a novel and unified cross-layer optimization framework, SFOptSec, specifically designed for LoRaWAN-based IoT environments. Unlike conventional studies that treat communication efficiency and data security as separate goals, SFOptSec holistically integrates adaptive spreading factor (SF) optimization with AES-128-based encryption modeling within a lightweight and energy-aware architecture. The framework dynamically balances energy consumption, reliability, and security by embedding cryptographic overhead directly into the optimization process, achieving real-time adaptation to changing network conditions. A multi-objective optimization model minimizes the combined cost of transmission and encryption energy, subject to reliability and security constraints. Furthermore, a feedback-based gateway adaptation mechanism iteratively refines SF and encryption parameters using observed metrics such as energy, latency, and packet delivery ratio (PDR). Simulation results demonstrate that SFOptSec significantly enhances network efficiency, reducing overall energy consumption and delay while maintaining robust data confidentiality. By unifying physical-layer optimization, link-layer encryption, and feedback-driven adaptation, SFOptSec establishes a scalable, secure, and energy-efficient solution for next-generation low-power IoT communication in dynamic LoRaWAN deployments.
Internet of Things (IoT) includes a group of connected, versatile devices. In this vast and intricate network of devices, Fog computing (FC) is becoming increasingly important as it helps manage the data flow of the networks. Technology for load balancing (LB) with effective resource allocation (RA) can be utilized to save energy usage and increase overall performance. Consequently, the primary emphasis now is on designing LB approaches to edge and fog scenarios. This research proposes a new model for RA and LB in fog computing (FC). Initially, RA in FC is carried out after getting the tasks from IoT devices. The tasks will be optimally allocated in each layer of FC based on the services. The resources are optimally allocated using a novel KBI-STO algorithm, considering constraints such as makespan, execution time, resource utilization, and MIPS. After the optimal allocation of resources, load balancing is performed optimally via the Kookaburra Integrated Siberian Tiger Optimization (KBI-STO) algorithm under specified migration constraints like migration cost and migration efficiency. Moreover, the proposed KBI-STO algorithm is a combination of the Kookaburra optimization algorithm and the Siberian Tiger Optimization algorithm, in which the innovation lies in the position updation. These enhancements efficiently explore novel areas and prevent early convergence. Finally, the experimental outcomes demonstrate the superiority of the KBI-STO algorithm over existing algorithms for optimal RA and LB in FE, in terms of execution time, makespan, etc., by varying the tasks. Furthermore, the suggested KBI-STO algorithm achieves minimal execution time, below 110 s, for varied tasks and varied virtual machines when compared to existing methods like STO, KOA, PFO, JSO, TSO, MPSO, and ACO algorithms.
In recent years, using “Electric Vehicles (EVs)” in an automobile zone has reduced the impact of air pollution and greenhouse gas emissions on environmental systems. Efficient charge management of EVs plays a significant role in improving vehicle safety, maximizing battery life and minimizing cost, which impacts the energy consumption ranges of EVs in the transportation system. Further, the batteries, along with the strong non-linear characteristics as well as time-variables, have been impacted by the random factors like operational conditions, and driving loads in the applications of EVs, and therefore, it is significant to tackle the limitations of the conventional “State of Charge (SoC)” assessment methods. This approach developed an innovative neural network learning methodology to solve an existing issue. In a suggested EV and its SoC estimation technique, essential temperature, available and requested battery thermal factor, current, actual power loss, air cooling temperature data, and power in a training process are aggregated through 10 different drive cycles. The acquired data is then provided for the feature extraction process. The conventional feature extraction models, like the Principal Component Analysis (PCA), have slow convergence in the training process, and it becomes complex for balancing feature selection and extraction performance, leading to overfitting issues. Thus, the developed method used the Restricted Boltzmann Machines (RBM)-based feature extraction mechanism to learn and capture high-level features from raw data in an efficient manner, and the extracted features are given to the charge estimation phase. The conventional Residual Long Short Term Memory (ResLSTM) approach does not enable safe and efficient discharging and charging performance, it reduces the lifetime of the battery, and is not capable of predicting the remaining range of energy in the EV. Therefore, the adaptive and attention mechanisms are incorporated with a ResLSTM in the developed work to produce the novel approach named Adaptive Residual Long Short Term Memory with Attention Mechanism (A-ResLSTM-AM) to significantly optimize the power consumption range and enhance the charge estimation process, which improves the battery life by avoiding overcharging and over-discharging. To further enhance the performance, the Hybrid Position of Tasmanian Devil and Black Widow (HP-TDBW) is used in the A-ResLSTM-AM for tuning the parameters. The traditional Tasmanian Devil and Black Widow (TDBW) struggles to quickly converge to the optimal solution in complex problems and cannot manage large datasets in a limited duration, to reduces the generalization ability. Unlike TDBW, the proposed HP-TDBW provides a novel contribution to efficiently tune the parameters of A-ResLSTM-AM. This optimization process can lead to reduced fuel consumption and energy losses during transmission and distribution. Hence, the implemented “EV SoC estimation technique” secures a superior presentation rate over other existing techniques in different experimental observations. It attains 34 in MEP, 0.25 in SMAPE, 114 in MASE, 5.2 in MAE, and 5.8 in RMSE measures in dataset 1. Also, it achieves 31 in MEP, 0.36 in SMAPE, 118 in MASE, 4 in MAE, and 5.1 in RMSE in terms of Dataset 2 validation. The proposed Adaptive Residual LSTM with Attention improved SoC estimation (RMSE reduced from 5.6
The convergence of cyber-physical systems, the Industrial Internet of Things (IIoT), and edge computing in Industry 4.0 has dramatically expanded the attack surfaces of industrial networks, making traditional intrusion detection systems (IDS) increasingly inadequate. While artificial intelligence (AI) and machine learning (ML) offer promising solutions, existing surveys often lack a specific focus on Industry 4.0 and a critical evaluation of the deployment feasibility. This systematic literature review (SLR) addresses these gaps through a PRISMA-guided analysis of AI-driven IDS research published between 2020 and 2025. From more than 8,000 studies, 22 high-quality papers were selected for detailed evaluation, revealing a pronounced shift towards edge-enabled detection architectures, hybrid AI models balancing accuracy and interpretability, and the integration of explainable AI (XAI) to strengthen operator trust. Key challenges persist, including reliance on synthetic datasets, limited validation in operational environments, computational demands unsuitable for resource-constrained edge devices, and integration issues with legacy operational technology (OT). The review’s contributions include a unified taxonomy mapping AI techniques to Industry 4.0 threats, a comparative analysis highlighting emerging trends such as federated learning and digital twins, and a research roadmap that emphasises lightweight models, realistic industrial datasets, and proactive autonomous response mechanisms. This SLR bridges the gap between academic innovation and practical deployment, supporting secure, intelligent manufacturing ecosystems.
Anomaly detection is a critical component for ensuring the security and reliability of Internet of Things (IoT) systems. Early detection of cyberattacks helps mitigate financial and operational risks for service providers. While many neural network-based anomaly detection models have been developed using centralized data, such centralized training introduces significant privacy and security vulnerabilities. Federated Learning (FL) has emerged as a promising paradigm to overcome this limitation by collaboratively training models without exchanging raw data. However, FL systems remain highly vulnerable to data poisoning attacks, where compromised IoT clients inject corrupted samples or manipulated labels into local training data, leading to degraded global model integrity and unreliable anomaly detection performance. To address this problem, we propose a robust and attack-resilient framework named Federated Learning with Shrink Denoising AutoEncoder (FL-SDAE), specifically designed to defend against data poisoning attacks in federated IoT environments. The proposed SDAE locally compresses data into a shared latent space and reconstructs it from corrupted inputs, allowing poisoned clients to exhibit higher training loss values. Leveraging this property, we introduce a novel Loss-Aware Aggregation (LAA) mechanism that adaptively identifies and filters out malicious client updates based on their abnormal training loss behavior during the aggregation process. Comprehensive experiments conducted on five benchmark IoT datasets (N-BaIoT, CICIDS, NSL-KDD, Spambase, and CTU13-08) demonstrate that FL-SDAE achieves higher anomaly detection accuracy and stronger robustness against both dirty-label (label flipping) and clean-label (Gaussian noise) attacks. Furthermore, comparisons with state-of-the-art defenses (Krum, Multi-Krum, Trimmed Mean, FoolsGold, and FLTrust) show that FL-SDAE consistently enhances model stability and resilience. Overall, FL-SDAE provides an effective defense framework that strengthens the robustness of federated anomaly detection models against diverse data-level poisoning threats in realistic IoT environments.
Vehicle-to-Everything (V2X) communications enable vehicles to exchange real-time information with other vehicles, infrastructure, pedestrians, and networks, playing a critical role in enhancing traffic safety, efficiency, and cooperative perception. As V2X deployments grow in density and complexity, congestion in shared wireless channels becomes a significant challenge, threatening the timely delivery of safety-critical messages. Although extensive research has explored congestion control mechanisms for V2X networks, existing surveys are fragmented, relying on inconsistent traffic models, evaluation metrics, and deployment assumptions, thereby impeding meaningful comparisons and comprehensive design insights. This review addresses this gap by providing a unified, systematic analysis of state-of-the-art congestion control mechanisms in V2X communications. The primary contributions of this paper are a comprehensive review and classification of congestion control strategies based on protocol layers, congestion detection and avoidance methods, decision-making architectures, and adaptability factors. It further presents a comparative analysis of existing mechanisms against key performance metrics and introduces a unified benchmarking framework alongside concrete design guidelines for future V2X congestion control development. A rigorous systematic literature review was conducted by applying clear inclusion and exclusion criteria to 35 peer-reviewed studies published from 2016 to 2025. The findings reveal that hybrid, cross-layer, and AI-driven congestion control mechanisms consistently outperform traditional rule-based methods in achieving scalability, fairness, and low latency, underscoring their suitability for next-generation V2X systems.
This research work focused on an attempt to quickly detect the vehicle crash on a road and eventually alarming the vehicles approaching towards the crash site. The purpose is to stop the subsequent collision of the fast-approaching vehicle in a challenging situation of poor visibility due to fog. The solution proposed to the problem is an integration of embedded system and sensor network encapsulated into a road stud. Each road stud are designed to sense the crash and signal the event to other studs to alert the vehicles coming on both direction of road. This research work considered factors such as the coverage area for sensing, power efficient methods and a reliable communication protocol. The coverage area is addressed for straight road and curvatures. The Power management is attained through adopting a proper Sleep/idle/Active switching mechanism and along with a low-power-long-term use strategy. A modified Sensor-Medium Access Control (S-MAC) is implemented in this proposed work. The system is tested for sensing and attained an accuracy of 97
The growing demand for energy, coupled with the increasing need for environmental protection and energy efficiency, is more pressing than ever. In this context, energy harvesting in networks offers a promising solution to reduce energy waste by utilizing ambient energy to charge devices. With the advent of 6G, new architectures have emerged, including data and energy integrated networks, which use radio frequency energy signals for energy harvesting. While the theoretical foundation of data and energy integrated networks (DEINs) has been established, their evaluation and verification is still difficult to ensure, and the analysis of their effectiveness and performance remains lacking. In this paper, we present a model-based approach utilizing UPPAAL to model, analyze, and evaluate the performance of DEIN networks. We explore various model configurations and conduct stochastic model checking to assess system behavior. Furthermore, we calibrate the model using real-world measurements to validate its effectiveness and demonstrate its capability to investigate diverse energy-related aspects.
Smart street lighting system (SSLS) is an intelligent outdoor lighting system with automated controls that enhances energy savings, safety, and urban planning. The need for better management and cost reduction for SSLS has motivated the development of new approaches and architectures based on the internet of things (IoT). Remote and autonomous control are prominent features of IoT-based systems that may improve street lighting operation. However, techniques for assessing availability, performance, and energy consumption of smart street lighting systems are not common. This paper presents an approach based on stochastic Petri nets (SPN) and reliability block diagrams (RBD) for assessing smart street lighting systems’ availability, performance, and energy consumption. Experimental results demonstrate the practical feasibility of the proposed approach. The system availability reached up to 99.74
The social internet of things (SIoT) represents an evolution of the internet of things (IoT), where smart devices operate as social entities, enhancing interactions and collaborations. However, ensuring trustworthiness among these devices poses significant challenges. This study introduces a decentralized incremental trust model designed to identify malicious nodes exhibiting dynamic behaviors, thereby improving the security and reliability of the SIoT ecosystem. The model employs incremental machine learning (ML) techniques to analyze device behavior and utilizes a fuzzy logic approach to evaluate service quality, taking into account both intention and capability. Additionally, our proposed model addresses the resource limitations of IoT devices by leveraging the benefits of a hybrid architecture. Validation through experiments on a simulated SIoT network demonstrates the model’s effectiveness in mitigating dynamic malicious activities, resulting in a significant improvement in attack detection rates.
The combination of the Internet of Things (IoT) with Software Defined Networking (SDN) technologies poses issues in terms of security and scalability. There are serious security vulnerabilities associated with traditional SDN systems since centralized controllers control them and are subject to manipulation by adversaries. In response to these issues, this paper suggests a unique method for automatically establishing security protocols in SDN settings by combining a Dynamic Trust Management Framework with the state-of-the-art reinforcement learning system Soft Actor-Critic (SAC). Our approach adjusts policy priorities dynamically depending on the reliability of network entities in real-time, improving security measures by using SAC’s capacity to learn optimum rules from data and adapt to changing network dynamics. To ensure authenticity and define user-specific attributes for access control, authentication methods are enforced throughout the registration process for both users and applications. With the use of SAC, security rules are created with permission activities, factual information, and temporal considerations in mind, strengthening network defenses against possible attacks. Moreover, policy conflicts are reduced by validation and storage in a centralized database, which streamlines administrative work. Our suggested model outperforms current approaches in performance assessments that are carried out via extensive metrics analysis and simulation using the iFog Sim tool. The outcomes of the simulation demonstrate how well our method works to improve the security and scalability of SDN, which is a major step forward for the security of networks enabled by the Internet of Things.
