CYBERSHIELD: HARNESSING ENSEMBLE FEATURE SELECTION TECHNIQUE FOR ROBUST DISTRIBUTED DENIAL OF SERVICE ATTACKS DETECTION
As Modern-day data increases in terms of dimensions (Instances X attributes), Single Feature Selection Techniques (SFST) often have certain biases and fail to provide optimal performance in machine learning models. To overcome the challenges associated with using SFST on large datasets, an Ensemble Feature Selection Technique (EFST) is proposed that includes multi-filter-based feature ranking selection and a wrapper-based feature subset selection methods for feature extraction on the Canadian Institute for Cybersecurity (CICDDoS2019) dataset. The EFST is a combination of three feature selection techniques viz; Chisquare, Recursive Feature Elimination (RFE), and Information Gain (IG), and the selection of features is based on the three conditions for aggregation. The EFST is then validated using the eXtreme Gradient Boost (XGBoost) classifier and the performance is compared with the individual feature selection techniques. This study reports that the proposed EFST outperforms the individual feature selection techniques in detecting Distributed Denial of Service (DDoS) with an accuracy of 99.81%.