Advanced Anomaly Detection Framework Using CNN-Grid Autoencoder Integration and Recursive Fuzzy Feature Selection Approach

Network intrusion poses a significant threat to data privacy and security in shared networks, often leading to cyber-attacks that compromise both system integrity and user data. Today's digital environment is seeing an increase in unauthorized actions, such as credential theft, illegal access, and data tampering. Despite the availability of various intrusion detection methodologies, accurately identifying and mitigating such threats remains a critical challenge. To address this issue, this paper proposes an automated attack classification model designed to enhance classification accuracy while minimizing errors based on input parameters. The proposed approach presents an innovative system for detecting network intrusions based on deep learning that integrates a fuzzy optimization method. The methodology begins with data pre-processing, including data cleansing and temporal smoothing, followed by feature extraction using Grid Convolutional Neural Networks (Grid-CNN). Optimal features are selected through a recursive multi-level fuzzy optimization algorithm. Finally, attack classification is performed using the SoftMax layer of the Grid-CNN architecture. The model is evaluated on the UNSW_NB15 dataset, with performance metrics including accuracy, precision, recall, and F1 score. Experimental results demonstrate that the proposed model achieves an accuracy of 96.02%, outperforming existing models in terms of both accuracy and robustness. This study highlights the potential of deep learning and fuzzy optimization in enhancing network intrusion