SSDP DDoS Attacks Detection using Naïve Bayes Classifiers with Wrapper Feature Selection Methods

One of the most destructive cyber-attacks nowadays is Distributed Denial of Service (DDoS). A Simple Service Discovery Protocol (SSDP) DDoS attack is established by using the advantage of holes in the Simple Service Discovery Protocol to overwhelm a victim network. Early SSDP DDoS attack detection is essential for minimizing consequences. The SSDP DDoS attack detection is analyzed in this work using various Naïve Bayes classifier types, including Gaussian, Bernoulli, Multinomial, and Complement. For feature selection, this study used various wrapper methods such as Forward Feature Selection, Backward Feature Elimination, and Recursive Feature Elimination. The CIC-DDoS2019 dataset is used in this study. The assessment measures used in this study were accuracy, K-Fold Cross-Validation accuracy, ROC-AUC, and log-loss values. Bernoulli Naïve Bayes classifier gives best accuracy value 99.96 with all feature subsets.