DoS Attack Detection Using Feature Selection with Information Gain and ML Classification
Supriya Vishal Dicholkar, Jagannath H. Nirmal · 2024
Designing an Intrusion Detection System (IDS) for attack detection in IoT networks is done by applying different machine learning and deep learning approaches. Standard CIE-CICIIDS2018 is a dataset with enormous benign traffic and different attack traffic with 80 features, which is used by most researchers. Feature selection or reduction is an important step in developing IDS, as an optimized featured dataset will give good results and attack detection accuracy in less time. In the present study, feature selection is carried out with the help of filter-based Information Gain in different batches, namely 20, 40, 60 and 78 features. After feature selection, two machine learning algorithms, namely K Nearest Neighbors (KNN) and Extreme Gradient Boosting (XGBoost), are applied. Based on the F1 score, accuracy, precision, and recall, a performance comparison is carried out for DoS Goldeneye and DoS Slow Loris attack detection. XGBoost outperformed to produce the greatest results when compared to KNN. Also, with 20 features and XGBoost model results are better as compared to KNN with all 78 features.