Performance Evaluation of Significant Feature for Interest Flooding Attack Detection on Named Data Networking

One of the internet architectures of the future that has advantages over the current system is Named Data Networking (NDN). However, Denial of Service (DoS) attacks, such as interest flooding attacks (IFA), can still disrupt the network. Detecting IFA attacks is crucial for preventing further damage. Several approaches to detection systems have been proposed, including a classification approach to detecting attacks with multiple detection parameters or features. However, the many detection system features that can be extracted from the network result in longer computation times for the classification algorithms. This research focuses on enhancing the detection of IFA by evaluating the features of the detection system and identifying significant features to improve detection accuracy and reduce computation time. We employed various feature selection algorithms, including information gain, wrapper naive Bayes, gain ratio, and correlation-based feature selection (CFS). The selected features are tested to detect attacks using several classification algorithms, including naive Bayes, random forest, J48, and Bayesian network. Our proposed method found only three essential features for detecting IFA from 18 features available, resulting in better detection accuracy and increasing by 47.8% the time to build the model. This study enhances NDN security while reducing computational cost, making real-time attack detection more feasible.