Feature Selection-Based Evaluation for Network Intrusion Detection System with Machine Learning Methods on CICIDS2017

Network attacks endanger the privacy and security of information. As a consequence of this, an effective mechanism is required for identifying network attacks. In this study, we show NIDS to increase classification accuracy by using the feature selection method, using various machine learning techniques, and comparing our model with various algorithms to improve features, a filter method as information gains importance is used and classified with machine learning methods like DT, RF, NB,LR, and KNN, and comparative analysis shows among algorithms. The simulation demonstrates that the proposed system outperforms baseline accuracy, precision, recall, and F-measure methods. For the CICIDS2017 dataset, the detection rate and false alarm rate of various forms of assault were higher than average. The analysis shows that our analysis has an impressive level of accuracy in classifying the NIDS with the CICIDS2017 dataset. Among all algorithms, Random Forest performed best with 95% accuracy and 93% F1-score.