A Meta-evaluation of Machine Learning Techniques for Detection of DDoS Attacks

Distributed Denial of Service Attack (DDoS) is a dynamic challenge in the field of network security. These attacks ban legitimate users from utilizing network resources as per their requirements. Intrusion Detection Systems (IDSs) can detect attacks up to a specific limit so it should always be equipped with a new type of defence solutions to combat the latest attacks. In this paper, authors evaluate the performance of various ML classifiers such as BayesNet, Naive Bayes, J48 and Random Forest to detect DDoS attacks. In this methodology, KDDCup99 data set is used for training and testing purpose. Principal Component Analysis (PCA) method is utilized for feature selection, choosing the most optimal features from the data set. By selecting top-ranked 20 features through PCA method, 10 fold cross-validation is done to measure the system's robustness. WEKA machine learning workbench is used to classify various attack types and validate its performance.