Comprehensive Intrusion Detection and Classification Using Deep Learning Techniques With Preprocessing and Feature Extraction

Abstract

For efficient identification of intrusions, this paper suggests computing techniques like recurrent neural networks (RNN), k-nearest neighbors (KNN), and convolutional neural networks (CNN) for classifying and predicting intrusions. Min-max scalability is used in preprocessing to normalize mathematical properties and guarantee consistency at various degrees. Linear discriminant analysis (LDA) extracts characteristics to increase the capacity for raw information discrimination. In addition, an innovative fusion of LDA and Min-Max scalability is investigated to maximize the depiction of features. Using CNN with extracted and feature-extracted data, this investigation expands the analysis to use the spatial organization of the convolutional CNN layers record. The tool used is Jupyter Notebook, and the language used is Python. Experiments on an incursion dataset show that the suggested mix of CNN, LDA, and Min-Max scaling operates dependably better than any of the distinct approaches regarding accuracy, precision, and recall.