ABSTRACT The existence of Bushfire is a significant problem. So not only are environmental threats to life and society a risk, but also the threatening of the safety of all. Thus it is extremely critical to detect these fires in time. This identification can assist in the successful application of fire management. Existing approaches might not be effective under conditions such as bad weather. A decrease in visible light also can affect accuracy. These techniques focus on the identification of fires during early stages. We introduce an innovative method in this study. The early detection of bushfires is achieved through UAVs. Each drone in the network has a thermal image camera onboard to provide real‐time surveillance of the assigned area. The drones are also equipped with sensors that detect smoke and other indications of potential fire. A control network receives up‐to‐the‐minute information from drones. Machine learning algorithms are used to interpret this data. The aim is to detect potential fires. The above detailed evaluations were captured using the eBee X and DJI Matrice 300 RTK. This study was carried out in a dedicated area provided by a National Park and involved the use of fixed and multirotor UAVs. State of the art sensor system could be successfully deployed on board the UAVs. The integrated technology was comprised of a Sequoia+ multispectral camera, H20T Quad Sensor and PGS 813 Gas Sensor. These were test burns, including controlled prescribed burning and fire simulations. They highlighted the system's impressive ability to detect bushfires in their early stages. The system has provided a quantum leap in preventive fire management. And it has heightened cost‐mindedness in real‐world settings. This enhancement is due to its high‐quality monitoring and precise detection of fire signs. The proposed method might be a useful and powerful tool for proactive fire fighting and early warning, such as fighting wildfire and related scenarios.

Fault detection in the Tennessee Eastman Process (TEP) is challenged by transient dynamics, post-injection regime transitions, and score drift that destabilize fixed-threshold detectors and increase false alarms. In the proposed Risk-Budgeted Conformal Anomaly Detection (RBC-AD) framework, a calibration-based anomaly detector integrates a causal forecasting backbone, frequency-aware scoring, and conformal decisioning. The forecaster is a dilated temporal convolutional network with causal self-attention, trained on fault-free runs to predict one-step-ahead Gaussian parameters by minimizing negative log-likelihood. Detection then scores each window using a fused nonconformity measure that combines the window-averaged predictive log-likelihood with an FFT-based frequency deviation term, using robust normalization fitted on a held-out calibration split and tuning the fusion weight under leakage-controlled splits. Conformal quantiles at miscoverage level α=0.05 define normal, uncertain, and anomaly states with finite-sample correction. Run-level partitioning over 500 runs per split and post-injection sampling (3,000 normal windows, 8,000 fault windows) achieves ROC-AUC = 1.000 and PR-AUC = 1.000 under the post-injection-only protocol; full-stream results provide the deployment-relevant assessment. Static conformal thresholds (θu=0.1977, θa=0.3033) yield F1 = 0.9989 and F1 = 0.9999. Deployment selects an uncertainty threshold to satisfy an empirical normal alarm-rate target of 0.05. Full-stream evaluation reports uncertainty rate and detection delay under k-consecutive alarm logic, with optional drift-triggered recalibration using normal-stream scores.

Due to technological advancements, today’s world is rapidly changing from traditional and legacy approaches to smarter and more autonomous techniques. Intelligent systems and super-connectivity make possible the era of pervasive services, where these services can be used by anyone, anywhere, and at any time (AAA). These advancements have completely transformed modern living, increasing convenience and providing greater comfort. Robotic cleaning is one such revolutionized area where self-driving robots and automated systems are used to navigate and clean dirt. 3D mapping adapts changes based on visual conditions to detect and remove waste obstacles from the location. These robots are useful for collecting waste and cleaning the floor. In this research work, we have proposed a LiDAR-based cleaning robotic system called ”CleanAll: Efficient Object Collection Route Planning from LiDAR Point Clouds in the Cleaning Scenario.” It is an automotive system that calculates an optimized path from its 3D point-cloud data to collect some objects through its object collector. CleanAll moves over the optimized path using point cloud data along with a dual-antenna GNSS receiver and an IMU. A defined data set is already used to compare and detect object shapes, and an extended Kalman filter is used to obtain the robot’s position. CleanAll is more responsive, and its quick decision-making makes it a better option for cleaning scenarios where dirt and trash are scattered throughout areas. The optimized path used by the robot is not only useful in finding the shortest path, but it also saves time and consumes the minimum energy by following these short paths for trash collection. CleanAll is evaluated in three different scenarios, collecting all objects (scattered/line-by-line) with an average standard error rate of 0.4593. To assess performance on other parameters, it has been tested with two standard datasets, ACIN and IPA. The TPRs have 0.75, and the FPRs have been recorded as 0.40 in ACIN, while the TPRs in IPA have 0.7 and 0.4 in the same scenario. This means that CleanAll performs better using the ACIN data set, achieving higher TPRs with lower FPRs. All these experiments illustrate that CleanAll performs better using the ACIN data set, which aims to recognize and clean trash more precisely and efficiently.

ABSTRACT A large amount of data have been published online in textual format for the last decade because of the advancement of information and communication technologies. This is an open challenge to organize and classify large amounts of textual data automatically, especially for a language that has limited resources available online. In this study, two types of approaches are adopted for experiments. First one is a traditional strategy that uses six (06) classical state‐of‐the‐art classification models (1. decision tree (DT), 2. logistic regression (LR), 3. support vector machine (SVM), 4. k‐nearest neighbour (k‐NN), 5. Naive Bayes (NB), and 6. random forest (RF)) along with two (02) ensemble methods (1. Adaboost and 2. gradient boosting (GB)) and second modeling technique is our proposed voting based ensembling scheme. Models are trained on a 75‐25 split where 75% of data is used for training and 25% for testing. The evaluation of the classification models is carried out based on accuracy, precision, recall, and F1‐score indexes. The experimental outcomes witnessed that for the traditional approach, gradient boosting outperformed for the limited resource language with 98.08% F1‐score, while SVM performed better (97.34% F1‐score) for the resource‐rich language.

ABSTRACT The continuous increase in news over the social web makes it difficult for users to search for a specific topic of interest. To overcome this difficulty, News Recommender Systems (NRS) are widely used to help users access the right and relevant content. Due to the diverse nature of NRS, researchers have started applying various deep learning (DL) models for the implementation of NRS. To our knowledge, however, current NRS do not look into the click‐through rate (CTR) with DL models. This manuscript proposes a news recommendation method that considers click‐log features with fine‐tuned DL models such as Artificial Neural Network (ANN), Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), Long Short‐Term Memory (LSTM), and Gated Recurrent Unit (GRU) for news recommendation. The primary objective of this study is to recommend news to users by constructing a click‐rate matrix from the provided datasets. In this regard, the datasets are preprocessed, and then a user‐item matrix based on click‐through rate is constructed. To demonstrate the applicability of the click, log feature matrix with deep learning models for news recommendation, extensive experiments are conducted using two cross‐validation methods, namely K‐fold and holdout. We used precision, Area Under the Curve (AUC), and Normalized Discounted Cumulative Gain (nDCG) to test the performance of the proposed method. According to the results obtained, CNN performed better compared to other DL models. The evaluation results displayed that incorporating the click log feature matrix and DL models can significantly improve the news recommendation results. The study also suggests that the use of the click log feature matrix with DL models is the most effective and simple method for recommending news articles, outperforming the existing approaches.

ABSTRACT Facial masks are still a big problem for regular facial recognition systems, especially in places where security is very important. This is because so many people wear them for health, cultural, or security reasons. This study presents a multi‐stage face reconstruction system utilising generative adversarial networks, aimed at restoring occluded facial regions while maintaining identity, structural accuracy and privacy protection. The suggested method uses gender categorisation, facial landmark recognition and mask segmentation to help with a landmark‐aware inpainting procedure. Separate training trajectories for male and female faces, as well as structural priors, help make reconstructions that are more accurate and consistent with their attributes. The model's main part is an encoder‐decoder generator that was trained with a composite loss function that balances perceptual quality, adversarial realism, pixel‐level precision and semantic coherence. The method takes a biometric privacy approach, rebuilding only the facial areas needed for recognition and hiding individually identifiable or unnecessary features to protect both recognition accuracy and user privacy. We built a huge matched dataset of 70,000 masked and unmasked face images from FFHQ to use for training and testing. The suggested strategy outperforms state‐of‐the‐art inpainting techniques, as shown by quantitative findings on various common metrics, such as structural similarity index (SSIM) (0.95), PSNR (33.3 dB) and identity similarity. In addition to technological contributions, our study moves forward the creation of AI systems that are ethical and open for use in sensitive areas like surveillance, border control and other areas where security and user privacy must be carefully balanced.

The significant increase in the number of Internet of Things (IoT) devices in various domains requires robust and adaptive authentication mechanisms. Existing methods often fail to address the dynamic and heterogeneous nature of the IoT ecosystem, resulting in significant security vulnerabilities. This paper presents a context-aware LLM-enhanced authentication framework (CLAF-IoT) that dynamically adjusts authentication protocols based on real-time environmental and user-specific contexts. Using the advanced contextual understanding and generation capabilities of Large Language Models (LLMs), the proposed framework enhances both security and usability in highly dynamic IoT environments. Key components include environmental context sensing, user behavior analysis, adaptive authentication protocols, real-time threat detection, and federated learning integration for continuous improvement and privacy preservation. Experimental evaluations demonstrate that CLAF-IoT achieves higher authentication accuracy in different scenarios, 11.11% false acceptance rate and 9.09% false rejection rate.

SUMMARY In Model‐Driven Development (MDD), detection of model defects is necessary for correct model transformations. Formal verification tools and techniques can to some extent verify models. However, scalability is a serious issue in relation to verification of complex UML/OCL class diagrams. We have proposed a model slicing technique that slices the original model into submodels to address the scalability issue. A submodel can be detected as unsatisfiable if there are no valid values for one or more attributes of an object in the diagram or if the submodel provides inconsistent conditions on the number of objects of a given type. In this paper, we propose a novel feedback technique through model slicing that detects unsatisfiable submodels and their integrity constraints among the complex hierarchy of an entire UML/OCL class diagram. The software developers can therefore focus their revision efforts on the incorrect submodels while ignoring the rest of the model. Copyright © 2013 John Wiley & Sons, Ltd.

Model errors are a major concern in the paradigm of Model-Driven Development (MDD) because of model transformations and code generation. It is important to detect model errors before transformation as in the later stages it is harder to trace and fix such errors. Formal verification tools and techniques can check the correctness of a model, but their high computational complexity can limit their scalability. In this research, we present a tool named UMLtoCSP (UOST) that uses a UML/OCL Slicing Technique (UOST) to verify complex UML/OCL class diagram. The tool accepts UML class diagrams annotated with OCL constraints as input, breaks the original model m into m1, m2, m3,...,mn sub-models while abstracting unnecessary model elements.

Model defects are a significant concern in the Model-Driven Development (MDD) paradigm, as model transformations and code generation may propagate errors to other notations where they are harder to detect and trace. Formal verification techniques can check the correctness of a model, but their high computational complexity can limit their scalability. In this paper, we consider a specific static model (UML class diagrams annotated with unrestricted OCL constraints) and a specific property to verify (satisfiability, i.e., "is it possible to create objects without violating any constraint?"). Current approaches to this problem have an exponential worst-case runtime. We propose a technique to improve their scalability by partitioning the original model into submodels (slices) which can be verified independently and where irrelevant information has been abstracted. The definition of the slicing procedure ensures that the property under verification is preserved after partitioning.

In the recent decades most of the big organizations have adopted maturity driven process improvement efforts (MDPI). Most of these efforts have been inspired of maturity models like the CMM (capability maturity model). The maturity of an organizationpsilas processes is measured through its maturity level. An organization availed a high maturity level is considered more trustworthy In this competitive business era making software process improvement (SPI) happen is a challenge for small organizations. The statistics provided by Software Engineering Institute (SEI) for software community striving for SPI by using CMM/CMM Iindicates that a large number of companies fail to achieve their process improvement goals. SPI efforts have mostly been prolonged, expensive, and not often delivered the effects back to the organizations in the same dimension as investigations.We wonder WHY? This research paper investigates strengths and weaknesses of maturity driven process improvement (e.g. CMM). The case studies in extant SPI literature are studied and focus group is used for data collection.The study suggests that process improvement initiatives should be tailored addressing organizational needs instead of blindly pursuing maturity models prescriptions.Furthermore, it is suggested to have an inception phase prior to an SPI initiative to decide whether a maturity driven process improvement approach should be opted or an effect driven process improvement approach.

Interoperability in telemedicine system is one of the major concern in telemedicine health care system. It is difficult to design exact and flexible interoperable architecture in telemedicine which transmit data and exchange information between systems to systems. The Service Oriented Architecture (SOA) is playing major role in the development of such system which helps to exchange the information between similar and dissimilar telemedicine applications. Using SOA and external Web services, the issue of interoperability can be resolved. The aim of this paper is to describe the importance of SOA in telemedicine through distributed system architecture design and implementation,which is developed in .Net platform using external Web services. The architecture of telemedicine system which we have developed is comprises of three layers that are presentation layer, business logic layer and data layer.

Distributed web-based applications have been progressively increasing in number and scale over the past decades. There is an intensification of the need for security frameworks in the era of web-based applications when we refer to distributed telemedicine interoperability architectures. In contrast, Service Oriented Architecture (SOA) is gaining popularity day by day when we specially consider the web applications. SOA is playing a major role to maintain the security standards of distributed applications. This paper proposes a secure web-based architectural design by using the standards of SOA for distributed web application that maintains the interoperability and data integration through certain secure channels. We have created CRUD (Create,Read, Update, Delete) operations that has an implication on our own created web services and we propose a secure architecture that is implemented on CRUD operations.