An Enhanced Intrusion Detection System Using Combinational Feature Ranking and Machine Learning Algorithms

In this modern era, where technology is booming, its dark side is also increasing constantly. Hence, now-a-days, it has become a necessity to develop systems capable of identifying and then nullifying all the security threats imposed. Several Intrusion Detection Systems(IDSs) have been developed already to identify network security threats using the latest technologies such as machine learning and feature selection techniques. But each of the feature selection technique gives different results with each classification algorithm. Hence, in this paper, we present a novel approach where the combination of feature selection techniques which includes Pearson Correlation (PC), Information Gain (IG), ExtraTreeClassifier (ET), and Chi-Square tests are used to rank the features using the weighted average. Thereafter, the different subsets of features are implemented to Random Forest, Decision Tree, Naive Bayes, Logistic Regression, and XG-Boost giving us the best detection accuracy and precision of 99.97% and 0.99980 respectively. These results are better than other similar intrusion detection related works. Moreover, for the preparation of IDSs authors have been constantly using the datasets which are already available but not up-to-date which makes the new attacks unidentified and hence, decreases their system performance. Therefore, we have also prepared our up-to-date dataset, consisting of benign traffic and malicious traffic, which has been prepared by introducing encrypted network attacks.