Machine Learning and Deep Learning framework with Feature Selection for Intrusion Detection

Increases in the size of the network and associated data have been a direct effect of technological breakthroughs in the technology and communication areas. As a result, new types of assaults have emerged, making it more difficult for network security systems to identify potential threats. An intrusion Detection is a critical cyber security method that keeps track of the progress of the network's software or hardware. In order to keep up with the ever-increasing rate and diversity of cyber threats, researchers have turned to machine learning approaches to build intrusion detection systems (IDS). Using machine learning algorithms, it is possible to identify with high precision the major differences between normal and abnormal data. In this paper, we proposed three feature selection techniques followed by machine learning and deep learning for IDS. We collected two different datasets and used the ANOVA F-value based method, impurity-based feature selection, and mutual information-based techniques for identifying the best features. Later, we applied three ML algorithms K-NN, Decision Trees, Logistic Regression, and Deep Learning Feed Forward Neural Networks on two datasets and achieved an accuracy of 88%, 99.9% with feed forward neural networks. The results shown that our model performed well compared to conventional methods.