We have presented the adaptive consistency framework for globally distributed databases that uses a machine-learned policy controller to balance latency, throughput, and correctness under dynamic workloads. This approach has treated consistency as a tunable knob, guided by real-time observability, workload characteristics, and service-level objectives (SLOs). A lightweight supervisor has collected end-to-end latency, read/write latency distribution, and data staleness metrics, and has selected a consistency level (e.g., strong, bounded staleness, or eventual) at the operation granularity or per session. The policy has learned offline from historical traces and updated online via a safe incremental learning loop that avoids destabilizing the system. The objective of this research is the formalisation of adaptive consistency as a policy-optimisation problem with stability guarantees. A learnable controller that integrates latency, staleness, and throughput signals. Practical guidelines for deployment, monitoring, and safety are also provided. We have implemented the framework on top of a representative distributed database prototype and evaluated it under synthetic and real workloads, including flash crowds, skewed key access, and partial network partitions. The results show a reduction of up to 28.6 percent in tail latency (p95/p99) with controlled staleness deviation, and a 75 percent improvement in overall throughput under bursty conditions, compared to 20 percent with static consistency configurations. We have considered the organisational concerns, security requirements, and opportunities for integration with the current Database-as-a-Service (DBaaS) platform.
Infrastructure as Code IaC is the accepted approach to provision cloud infrastructure declaratively. Still, misconfigurations in IaC remain the primary cause of cloud security incidents, accounting for 67 percent of all disclosed cloud breaches. However, the current set of countermeasures, such as rule-based static scanners, policy-as-code tools, and manual review gates, is inadequate to prevent such misconfigurations or to address them through autonomous remediation. In this paper, the authors propose a multi-agent generative AI system called GenSecOps, comprising four agents that work together to prevent misconfigurations in IaC. These agents are the IaC Understanding Agent IUA, which uses the IaC artefact to create a semantic resource graph; the Risk Prediction Agent RPA, which uses a hybrid model of the Transformer and Graph Neural Networks to create risk mappings; the Generative Remediation Agent GRA, which uses the risk mappings to create corrected policy-compliant IaC templates; and the Autonomous Enforcement Orchestrator AEO. Experiments on a corpus of 48,000 IaC templates Terraform, CloudFormation, Kubernetes show that GenSecOps achieves a misconfiguration detection F1 score of 0.934, a 73.2 percent reduction in critical findings overrule based baselines, an 81.5 percent improvement in mean-time-to remediate MTTR, and drift-recovery latency below 4.2 minutes. These results demonstrate that generative AI agents provide a viable, deployable foundation for self-healing, autonomously secured cloud-native infrastructure.
As the internet-connected systems have expanded, albeit briefly, intrusion detection systems IDS have become an important element in cybersecurity. Conventional host-based and network-based IDS systems are unable to detect advanced and distributed attacks promptly. Collaborative Intrusion Detection Systems CIDS enhance precision by enabling nodes to share intelligence. Nevertheless, CIDS usually have issues associated with secure data sharing and trust. To overcome these constraints, this paper presents a smart, collaborative intrusion detection system based on the Ethereum blockchain. Decentralised trust, data immutability, and the absence of a central authority are guaranteed through blockchain integration. Both the signature-matching and fuzzy genetic algorithms are machine learning algorithms used in anomaly- and signature-based intrusion detection. Datasets such as NSL KDD, CIC IDS 2017, and CIC IDS 2018 are used to assess the systems performance. Findings show better accuracy, fewer false positives and greater resilience in the multi-node environments. The suggested architecture will include IDS tools such as Snort, Zeek, and Suricata, and will be combined with smart contracts to enable secure cooperation. The given work contributes to the development of the field by combining AI, blockchain, and CIDs to provide a new solution to the current threats posed by cybersecurity violations in distributed networks.
The combination of a low-voltage DC LVDC microgrid with an Intentional Islanding Algorithm IIA is presented in this paper to guarantee stable and reliable operation during microgrid disturbances. By isolating the LVDC network and identifying abnormal grid conditions, the proposed algorithm enables a rapid transition from islanded to grid-connected mode. Solar photovoltaic generation and battery energy storage are incorporated into the system to support sustainable energy use and maintain power balance during on-site operation. The performance of the proposed II algorithm is compared with conventional islanding detection algorithms, such as the passive islanding algorithm, the active islanding algorithm, and the hybrid islanding algorithm, using parameters including voltage stability, settling time, power balance, continuity of supply to critical loads, and power quality. A modified IEEE-recommended distribution system serves as the foundation for the LVDC microgrid model, implemented in MATLAB Simulink. Simulation results demonstrate that the proposed IIA significantly improves system performance by reducing voltage fluctuations, accelerating system stabilisation, improving DC-link current behaviour, and ensuring uninterrupted power supply to critical loads during grid outages. These results confirm the effectiveness of the proposed approach in improving the reliability and resilience of LVDC microgrids.
Skin cancer screening using dermoscopic images remains challenging because malignant and benign lesions can share overlapping visual cues, including irregular borders, heterogeneous pigmentation, and complex textures. In addition, real clinical datasets often contain ambiguous samples, acquisition artifacts, and non-uniform illumination, which may degrade training quality and generalization. Although deep learning has achieved strong performance in dermoscopic lesion classification, its dependence on large annotated datasets, limited interpretability, and computational demands can limit adoption in resource-constrained clinical settings. This paper presents a complete classical machine learning framework for four clinically significant dermoscopic categories: basal cell carcinoma (BCC), melanoma, nevus, and pigmented benign keratosis (PBK). A unified hand-crafted feature representation is constructed by combining texture descriptors (Histogram of Oriented Gradients and Local Binary Patterns), region geometry and border regularity measures, Hu moment invariants, lesion intensity statistics, and multi-space colour descriptors (RGB/HSV statistics and HSV histograms). To reduce the effect of label noise and hard-toseparate samples, a margin-based smart cleaning strategy removes low-reliability instances on a per-class basis using a preliminary classifier’s confidence margin. Class imbalance is then mitigated through controlled upsampling to a common class count without synthesizing new image content. Feature dimensionality is reduced using minimum-redundancy maximum-relevance (mRMR) ranking, retaining the top 300 features to balance accuracy and runtime. We evaluate both flat multi-class classification and a clinically motivated two-level hierarchical design that first separates Cancer vs non-Cancer, then performs subtype classification within each branch (BCC vs melanoma; nevus vs PBK). Experiments are implemented in MATLAB R2022 using stratified 5-fold cross-validation with strict fold-wise isolation to prevent information leakage. The best flat model, SVM with an RBF kernel, achieves a mean accuracy of 82.87%. The hierarchical system achieves an overall accuracy of 81.80%, with Level-1 accuracy of 86.54%. For Level-1 malignancy detection, ROC AUC = 0.9383 and PR AUC = 0.9367 (all folds combined), indicating strong discrimination. Oracle branch evaluations confirm high in separability (94.04% for BCC vs melanoma; 94.19% for nevus vs PBK) and show that residual loss is primarily due to Level-1 routing errors.
Accurately forecasting the breakdown voltage of insulating oils is a prerequisite for the reliable design and operation of high-voltage equipment. The present work focuses on developing data-driven artificial intelligence (AI) models to predict the breakdown voltage of transformer oil as a function of temperature and electrode spacing. Two different machine learning algorithms are applied and compared: Gaussian Process Regression (GPR) and Radial Basis Function (RBF) neural network. The experimental data for electrode distances of 5 mm and 20 mm are used to train, test, and validate the models using a 60/20/20 data-splitting scheme. The predictive capacity of the models is evaluated using the three metrics: mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R). Experimental results confirm that the model predictions are in excellent agreement with the measurements at short electrode distances for both models. Nevertheless, at longer distances, the differences between the two performances become quite substantial. The GPR method is more reliable and generalises better, particularly at 20 mm, where it yields lower validation errors than the RBF approach. In addition, as a probabilistic method, GPR enables the estimation of predictive uncertainty, which is essential for applications oriented toward safety and dependability. Overall, the present work has demonstrated GPR's capability to determine the breakdown voltage of insulating oils and its potential for high-voltage insulation diagnostics and design.
Nowadays, nine out of ten people inhale polluted air, causing dangerous health concerns. This means that air pollution poses a serious threat to societys health. It supports enhanced dimension, cause detection, prediction, expectation, and logical problem-solving. AI technology can rapidly and accurately detect air pollution. AI has quickly exposed the extent of air pollution. This study estimates logistic regression (LR) and Random Forest (RF) models, two widely used statistical methods for predicting long-term air pollution and environmental health. Logistic regression may predict air pollution more effectively than other machine learning approaches. The objective of this analysis is to improve the algorithm's performance during the collection activity and reduce air pollution. The average detection accuracy falls within one standard deviation, indicating that the proposed model is as efficient as, and more effective than, the modern method. Logistic Regression and Random Forest (which is valued the highest accuracy (0.93) and precision (0.92)).
Workers involved with underground pipelines might be subjected to dangerous transient voltages caused by adjacent lightning strikes on electrical towers linked to grounding systems. To better understand and predict these indirect effects, this paper investigates the transient electromagnetic behavior of a tower– grounding grid–pipeline system subjected to a direct lightning strike. The entire setup is simulated using transmission-line theory to enable a thorough understanding and accurate modelling of wave propagation, electromagnetic coupling, and ionisation across the various system components. A large-scale setup comprising five transmission towers linked by a grounding grid, and located near an underground pipeline with a total length of 2.8 km, is considered in this paper. The lightning current of 12. 5 kA is delivered at the first tower's top, and the pipeline transient currents and voltages induced are calculated. The lightning-wave propagation from the strike location through the tower arms, grounding grid, and soil to the pipeline is investigated under the assumption of uniform soil conditions. Different soil resistivities (100, 300, and 600 m) are used to assess their effects on system behaviour. Three electrical pipeline models, all based on transmission-line equivalences, are constructed and compared. The first model explores only resistive effects, whereas the second and third gradually incorporate inductive, capacitive, and conductive elements, thereby enabling a more precise depiction of electromagnetic coupling and dielectric losses. According to the simulations, the lightning current amplitude fades progressively through the resistive, inductive, and capacitive components of the tower, grounding grid, and pipeline. As expected, the induced current is maximum at the struck tower, and it decreases along the system. About the first pipeline model, the induced voltages were always at a level safe enough for personnel, regardless of the soil resistivity considered. However, the second and third models showed a significant increase in pipeline voltage, with the third model exhibiting very high voltages despite lower current magnitudes. Hence, the results clearly underscore the importance of pipeline modelling, soil resistivity, and electromagnetic coupling in evaluating lightning-induced hazards. The modelling approach introduced here not only advances understanding of the transient behaviour of grounding systems and pipelines subjected to lightning but also enables the development of safer grounding layouts, pipeline materials, and protective measures that better shield people from lightning hazards.
Automated machine learning (AutoML) has emerged as a practical approach to facilitate the adoption of machine learning by automating model development tasks, including preprocessing, model selection, and hyperparameter optimisation, thereby reducing reliance on specialised expertise. Recently, major cloud providers have integrated AutoML into their platforms to offer end-to-end machine-learning pipelines as managed services. However, the practical implications of cloud-based AutoML systems, particularly their system and operational aspects, remain insufficiently explored. This paper presents an empirical, system-oriented analysis of AutoML services provided by Microsoft Azure, Amazon Web Services, and Google Cloud Platform. Using representative regression and binary classification tasks, the predictive performance, evaluation metrics, and feature-importance results produced by each platform are compared. The study also examines how platform-level design choices influence usability, reproducibility, and lifecycle management. The results demonstrate that cloud AutoML platforms deliver high-performing models that operate without manual intervention, whereas differences among providers primarily reflect architectural and operational abstractions rather than algorithmic limitations. These findings suggest that cloud AutoML should be understood as an integrated system that combines automated modelling and MLOps capabilities and offers a viable pathway toward production-ready machine learning under real-world constraints.
There is a gap in industrial cybersecurity. Above all, the frameworks that govern OT security, IEC 62443 for short, were built for a time when the air gap existed, and the threat model was physical. That era came to an end quietly, between the first remote vendor access agreement and the first cloud-connected historian. By 2024, more than 12,000 ICS-oriented cybersecurity incidents had occurred in one year, with dual IT/OT breaches averaging USD 4.56 million per event. Zero Trust Architecture, as specified in NIST SP 800-207, is the correct conceptual response: No longer should you trust your network location; verify everything at every step. The caveat is that NIST SP 800-207 is IT-oriented, and its accompanying implementation manual specifically excludes OT. No Zero Trust standard is specifically designed for OT. This paper examines the top-level cybersecurity governance platforms and their relevance to five dimensions of OT, presents a scenario of 2024-2026 industrially harmful environments, and introduces the Adaptive Zero Trust Framework for Industrial Control Systems (AZTF-ICS). AZTF-ICS is an innovative five-pillar model that uses Zero Trust principles to address the unique operational constraints of mission-critical process automation, with real-time requirements, high availability, and safety tasks that are not susceptible to interruption, irrespective of any security control policy.
Various eye diseases appear differently in red, green, and blue channels of RGB colour models. Colour channels provide the primary information for detecting eye diseases. Selection and training of these channels are the primary tasks in pre-processing colour fundus images and the automatic detection of various eye diseases. Improvements in quality and appearance, as well as image enhancements, are performed during the pre processing stage without affecting the accuracy of fundus images. PSNR, MSE, DSSIM, FSIM, RMSE, UIQI, and SSIM are calculated to preserve structural information between the original image and colour-converted images. The test images are taken from the DRIVE fundus database and evaluated using colour-space structural models. The methods are tested using an OpenCV Python Jupyter notebook on a Windows platform with an Intel i5 processor at 3 GHz and 16 GB of RAM. The results are compared to determine the best colour space model for detecting cotton wool spots before post-processing.
The increase in harmonic distortion in power networks due to non-linear loads is now a major challenge for utilities, and utilities are finding ways to reduce these harmonics by imposing penalties on customers that generate them. This research aimed to design a fairer harmonic penalty model for customers who inject harmonics into power systems, accounting for harmonics injected from both the customer and utility sides. The harmonics injected from the customer and utility sides of the PCC (Point of Common Coupling) were evaluated, and the resulting Total Harmonic Distortions (THD) were used to assess the penalty and compensation due to the customer. The customers electricity bill was also evaluated using this penalty model. The application of this penalty model results in a fairer model in which utility-penalised customers are compensated for harmonics injected from the network that exceed the limits, as well as for harmonics injected from the utility side. This research offers a substantial improvement over existing penalty models, as customers were treated fairly by considering the harmonics generated by their networks and the utility networks.
In the context of augmented reality (AR), it is a revolutionary platform that enables the simulation of lively virtual experiences in a real-world setting. The project, called an immersive experience, imagines the dragon flight in the open and through narrow passages. One of the key features of the project is real-time path planning, which involves automatic detection and avoidance of dynamic obstacles, along with a user interface which considers control responsiveness and immersion. The project aims to engage the user by blending spatial awareness and responsive interaction within the virtual environment. One of the main goals is to create an intuitive and captivating simulation of virtual objects by means of breaking down spatial and occlusion limitations. The last-mentioned refers to human comprehension.
The energy optimisation of buildings relies heavily on an accurate understanding of the thermophysical properties of building materials. However, in many developing countries the characterization of these materials remains limited by the high cost and scarcity of measuring instruments. In this perspective, we first designed, realized and experimentally validated a device for the simultaneous measurement of thermal conductivity and diffusivity of building materials. With these results, the present study introduces the design and simulation of an autonomous electronic thermal data-acquisition device intended to improve the accuracy and autonomy of the initial system. The proposed device is based on an Arduino Due and integrates a temperature sensor (MCP9600), a real-time clock (DS3231), a 20x4 LCD, and an SD Card storage module. The simulation, performed in Proteus, validated the functional consistency of the circuit and the communication between the different modules via the I2C and SPI interfaces. The results demonstrate stable acquisition and reliable real-time temperature recording. This work aligns with efforts to develop affordable, robust, and reproducible measurement instruments tailored to the specific needs of African energy and materials science research laboratories.
Preserving Indian folk music in digital repositories poses significant challenges because robust classification systems are lacking to capture its linguistic, instrumental, and acoustic diversity. As a cornerstone of Indias intangible cultural heritage, this music faces the risk of marginalisation and loss unless systematic, scalable methods are employed to identify and preserve it. This research aims to develop an automated, multi-modal framework for regional classification of Indian folk music, thereby enabling structured archiving and improved accessibility. To achieve this, a novel machine learning pipeline was designed, integrating Whisper for speech recognition and regional language identification. Instrument detection was performed using YAMNet, which has proven effective in recognizing traditional instruments . Acoustic features such as MFCCs, chroma, and spectral descriptors were extracted using Librosa. Together, these tools provide a comprehensive understanding of the songs linguistic, instrumental, and rhythmic content. The curated dataset includes folk music from linguistically rich regions of India, such as Marathi, Punjabi, Urdu Qawwali, and dialects from Uttar Pradesh and Bihar. Seven supervised learning algorithms were trained and evaluated, including Random Forest, Support Vector Machine, and Gradient Boosting. Simpler classifiers, such as K-Nearest Neighbours, Naive Bayes, and Logistic Regression, were also tested. A hybrid ensemble model combining Random Forest, SVM, and Gradient Boosting through soft voting achieved a classification accuracy of 99 percent. This result demonstrates the effectiveness of ensemble learning, combined with multimodal features, in handling nuanced differences in regional folk genres. This research addresses the critical gap in scalable and automated tools for preserving folk music. The study highlights the potential of artificial intelligence in safeguarding endangered cultural assets.
The rapid expansion of complex system logs in modern infrastructures has heightened the need for accurate, interpretable, and low-latency risk analysis. These logs contain high-dimensional, context-rich data that is essential for operational reliability, cybersecurity, and compliance. While conventional machine learning models are efficient, they often overlook the nuanced semantic relationships in sequential log data, limiting predictive reliability. Conversely, large language models (LLMs) offer deeper contextual understanding but are computationally intensive, making them unsuitable for real-time, large-scale deployment. This study presents a deployment optimised pipeline that balances semantic depth with computational efficiency for log-based risk prediction. The architecture integrates lightweight MiniLM embeddings with an XGBoost classifier to produce interpretable, high-quality predictions at reduced computational cost. Key optimizations include class balancing to address dataset skew, model quantization to lower memory usage, and batched inference to increase throughput, enabling cost-effective CPU-only execution without GPUs. A structured evaluation examined accuracy, latency, and memory trade-offs across production scenarios. Testing on representative log datasets showed notable gains over a TF-IDF baseline: classification accuracy improved from 21.4 percent to 57.1 percent, weighted F1-scores rose accordingly, and inference latency decreased with negligible loss in predictive strength. By combining transformer-based dense embeddings with gradient-boosted decision trees, this approach delivers a practical balance of semantic expressiveness, interpretability, and deployment efficiency. The framework supports scalable, real-time risk prediction for cybersecurity monitoring, compliance auditing, and IT operations, bridging the gap between advanced language modelling and real-world infrastructure constraints.
Recently, approximate computing has become a wellknown computer outlook. It is a broad field with new research paths emerging daily. Approximate computing systems enhance energy efficiency and computational speed at the expense of precision in output. From a computational standpoint, this paper offers a concise and thorough overview of recent research areas and accomplishments in energy-efficient computing. We classify and analyse the machine learning techniques used in approximate computing applications. Approximate computing is used at the software, circuit, and hardware levels. Machine learning (ML) methods are crucial in various approximate computing applications, enabling performance improvements at multiple levels. The scope of the systematic literature review encompasses an in-depth examination of the most prominent machine learning (ML) trending techniques in approximate computing applications. This paper also addresses recent breakthroughs in approximate computing hardware, software, and approximate data communication.
LoRaWAN (Long Range Wide Area Network) contributes as a key technology in Industrial Internet of Things (IIoT). LoRaWAN receives consistent industrial attention due to its far-reaching long-range communication capability, low power consumption, and easy-to-adapt architecture, which requires minimal infrastructure costs. This paper provides more insightful information about LoRaWAN in IIoT, illustrating how LoRaWAN optimises operational efficiency, safety, remote data collection, and sustainability through its dynamic features in industrial settings. Additionally, this paper explores the integration of LoRaWAN with emerging technologies, including artificial intelligence, machine learning, edge computing, and blockchain. While this paper emphasizes the potential of promising trends in the IIoT ecosystem, it also explores relevant challenges like scalability, security, and interoperability that demand the need for continued research. The comparative analysis of recent research works highlights key technical parameters. It is helpful to critically examine the trade-offs between performance, scalability, and security, and address them with novel solutions. In summary, this paper offers valuable insights for industry professionals and researchers seeking to leverage LoRaWAN to advance the Industrial Internet of Things (IIoT).
This paper introduces a dynamic channel-assignment algorithm to minimize interference among Access Points (APs) in Wireless Local Area Networks (WLANs). The algorithm utilises Particle Swarm Optimisation (PSO) to efficiently optimise channel allocation, ensuring a positive Signal-to-Interference Ratio (SIR) and Signal-to-Noise Ratio (SNR) for all users, thereby guarantee ing reliable communication. The process commences with an ini tial channel assignment, followed by iterative refinement using PSO to reduce inter-AP interference. PSO is employed to accu rately compute SIR and SNR values for each user, thereby provid ing a precise assessment of signal quality. Experimental results demonstrate substantial improvements in both SIR and network throughput, confirming the algorithm's effectiveness in reducing interference. Furthermore, the proposed approach is adaptable to various WLAN scenarios, making it suitable for diverse user dis tributions and network loads, thereby ensuring wide-ranging ap plicability in real-world settings.
Concrete is a widely used construction material, but one of the significant problems is cracking in structures, which may later cause severe damage to the structure. An ultimate solution for this problem is the self-healing of concrete. This study employs an experimental approach to investigate self-healing in M25 grade concrete using silica fume and a superabsorbent polymer. This experimental approach utilised a fixed proportion of silica fume and SAP. The amount of silica fume was 10% of the cement weight, and 0.45% of SAP was added to increase the self-healing efficiency. The strength of the concrete specimen was also tested. It was found that concrete specimens made with silica fume, along with SAP, exhibit greater stability compared to concrete specimens made with silica fume alone. It was observed that crack closure in concrete made with silica fume took 33 days, while concrete specimens made with silica fume along with SAP took 17 days. This study demonstrates the potential of using silica fume and SAP in concrete to enhance the durability and lifespan of structures, thereby promoting sustainability.
