Critical Feature Selection (CFS) Techniques for Early DDoS Attack Detection in the Internet of Things Environment

The introduction of the heterogeneous IoT with the distributed denial of service (DDoS) attack is one of the severe security risk that has arisen in the digital world. The severe impact of these massive DDoS attacks has been demonstrated repeatedly by crippling various infrastructures, leading to major losses and compromising the overall accessibility of the digital world. Thus, addressing the risks associated with DDoS attacks has become a critical priority to mitigate their impact on the HetIoT infrastructure. The primary objectives of network security are to secure the sensitive data for the inherent application and to shield the information broadcasted over a network. We, therefore, propose implementation of Critical Feature Selection Technique for Early DDoS Attack detection in the IoT environment. The methods employed for feature selection are Pearson Correlation Coefficient, Information Gain, Random Forest Importance and Kendall's Rank Coefficient. This enabled the effective extraction of the most significant features, thereby enhancing the efficiency and accuracy of the analysis. Decision Tree, Random Forest, XgBoost, AdaBoost tree classifiers are utilized for the analysis. Ensemble was then used to produce an optimal predictive model which helped in increasing the accuracy of predictions, and providing robust results. The tree classifiers give 100% accuracy for binary classification of DDoS attacks.