An Effective Design of Intrusion Detection System With Classification Algorithms And Feature Reduction In Machine Learning

To safeguard networks from outside attacks, the Intrusion Detection System (IDS) has become an essential tool in the modern realm. With the proliferation of data-generating and -sharing tools like Big Data, CC, and the IoT, it has become more challenging to isolate characteristics that contribute to effective intrusion detection systems (IDS). This problem has been addressed by using feature selection techniques (FSA) to filter out superfluous characteristics and isolate critical ones from network data. This has resulted in the creation of intrusion detection system models that are suitable for large-scale networks and have reduced costs while producing greater performance. A number of FSA classifiers have been investigated and tested in an effort to identify the best one for intrusion detection system capture. On real-time datasets, the top-performing classifier utilizing an appropriate feature reduction approach (FST) has been tested. Separately, we compared the efficacy and performance of FSA classifiers to those that included all characteristics. The proposed model (DT + RFE) is said to enhance IDS performance on the NSL-KDD, CICDDoS2019, and CICIDS2017 datasets, achieving 99.21%, 99.97%, and 99.94% average accuracy, respectively, while decreasing computation cost.