As fraudulent transaction methods evolve rapidly; it becomes progressively more challenging to detect them in payment systems. Static machine learning and rule-based traditional detection methods cannot capture all the dynamic and evolving nature of fraudulent behaviors, resulting in lower detection accuracy and a higher false positive rate. This study proposes a complete framework that brings together advanced data preprocessing, effective drift detection, and a reliable detection model to address these issues. The method uses Mutual Information and SelectKBest for selecting important features, applies ADASYN to handle class imbalance, and adopts Convolutional Neural Networks (CNN) to capture complex transaction patterns. By implementing Early Drift Detection Method (EDDM) and ADaptive WINdowing (ADWIN), the drift can be detected in advance and the system can respond to changes, both gradual and sudden.The framework was evaluated on three datasets, including real-world transactions and mixed-data environments, achieving superior accuracy, precision, and drift detection rates, with values up to 99.99% accuracy and 1.0 respectively. The findings show that the framework can adjust to changing patterns of fraud, reduce false positives, and enhance detection performance. These insights demonstrate the significance of dynamic preprocessing and drift-aware approaches in the context of real-time fraud detection. This also serves as a basis for future work in adaptive fraud detection model research areas such as of integrating online learning for improved speed and efficiency in high-frequency transactional environments.

The adoption of cooperative intelligent transportation systems (cITSs) improves road safety and traffic efficiency. Vehicles connected to cITS form vehicular ad hoc networks (VANET) to exchange messages. Like other networks and systems, cITSs are targeted by attackers intent on compromising and disrupting system integrity and availability. They can repeatedly spoof false information causing bottlenecks, traffic jams and even road accidents. The existing security infrastructure assumes that the network topology and/or attack behavior is static. However, the cITS is inherently dynamic in nature. Moreover, attackers may have the ability and resources to change their behavior continuously. Assuming a static IDS security model for VANETs is not suitable and can lead to low detection accuracy and high false alarms. Therefore, this paper proposes an adaptive security solution based on deep learning and contextual references that can cope with the dynamic nature of the cITS topologies and increasingly common attack behaviors. In this study, deep belief networks (DBN) modeling was used to train the detection model. Binary cross entropy was used as a loss function to measure the prediction error. Two activation functions were used, Relu and Softmax, for input–output mapping. The Relu was used in the hidden layers, while the Sigmoid was used in the last layer to map the real vector to output between 0 and 1. The adaptation mechanism was incorporated into the detection model using a moving average that monitors predicted values within a time window. In this way, the model can readjust the classification thresholds on-the-fly as appropriate. The proposed model was evaluated using the Next Generation Simulation (NGSIM) dataset, which is commonly used in such related works. The result is improved accuracy, demonstrating that the adaptation mechanism used in this study was effective.

The increasing number of Internet-enabled devices has demonstrated the need to have accurate intrusion detection systems (IDSs). To address this, we adapt the structure of two-dimensional convolutional neural networks (2D CNNs). Particularity, we restructure the inputs and tune convolutional/dense layers (kernel sizes and activations) to non-image features. Using BoT-IoT 2020 dataset, we benchmark the accuracy, precision, recall, F1-score, AUC-ROC, and the false-positive rate (FPR) of a batch-size of 32-1024. The adapted 2D CNNs achieve a 99% accuracy and a lower FPR at a batch size 128, which suggests that it is efficient in detecting IoT-based attacks. In addition, we compared the adapted 2D CNN with powerful tabular baselines, LightGBM has lower FPR (4.8%) and higher AUC (94.8%) at detection accuracy (about 94.7%) whereas MLP/TabNet have high recall, precision and high FPR (about 19.26%), highlighting precision recall/FPR trade-offs.

Spinal cord diseases are prevalent worldwide and affect people of all age groups. Since manual diagnosis is time consuming, aids of Artificial Intelligence (AI) are involved in disease diagnosis. Several approaches have been developed for spinal condition classification. However, complexities occur due to computational overhead or misclassifications. Intending to introduce a novel disease diagnosis model, this paper presented a NeuroFusionNet to accurately detect various spinal cord diseases. For effective classifier performance, significant preprocessing is mandatory. For this reason, this research incorporates median filtering, Modified Local Contrast Mapping (M-LCM), and Contrast Limited Adaptive Histogram Equalization (CLAHE) to prepare the raw data for further processing. Besides, partitioning the spinal cord from the Magnetic Resonance Imaging (MRI) is necessary to enhance visualization and accurate diagnosis using proposed 3D Residual Multi-Scale Attention U-Net with Dilated Convolution and Squeeze-and-Excitation Networks (3DRMAU-Net-DCSE). Significant features are then extracted via Convolutional Autoencoder (CAE) and optimal features are selected via proposed Advanced Dwarf Mongoose Fox optimization (ADMFO). Finally, an innovative deep learning (DL)-based NeuroFusionNet is designed using Vision Transformer (ViT), Temporal Convolutional Neural Network (CNN)+, and attention-based Capsule Network (CapsNet) to enable reliable diagnosis and decision-making in spinal health conditions effectively. To evaluate the performance of the suggested model, a comparative study is conducted with benchmark models for several performance metrics such as accuracy, recall, and F1-score. On the spine MRI dataset, the stated DL model achieved an overall accuracy of 98.88%, sensitivity of 98.95%, and F1-score of 98.95% and proved its significance over other models.

In the realm of addressing the primary unfeasible issue in the cybersecurity in smart grids to ensures protection of critical infrastructure, a novel “Anomaly-Graph Fusion Isolation Framework (AG-FIF)” was introduced. In the existing, unauthorized access to smart meter data due to weak encryption and poor authentication in smart grid systems, while scalability challenges hinder real-time processing. Hence a novel, Integrated Anomaly Detection Ensemble (IADE) was introduced in this approach, offers strong defence against manipulation of scalability problems in smart grid settings with Support Vector Machine (SVM) for classification purpose, Random Forest for integration purpose and Autoencoders for reconstruction purpose. Moreover, the smart grid information networks are being swamp with DDoS attacks due to bad management of resources, while standard methods have a difficulty in detection and relief in dynamic distributed environment. Thus, a novel Graph-Enhanced Isolation Tree Framework (GEITF) was invented, which controlling network resources and lessening the effects of dispersed and dynamic DDoS attacks with Graph Convolutional Networks (GCNs) for modelling purpose, Isolation Forest for detection purpose and Decision trees for classification purpose. The AG-FIF framework delivers 98.55% accuracy, 97.35% F1-Score, low FNR (1.01%), FPR (1.03%), and excels in resource management, robustness, and scalability.