An efficient DDoS attack detection and categorization using adolescent identity search-based weighted SVM model

Abstract

Distributed Denial of service (DDoS) is the most frequent as well as serious attack that causes serious threats to new-generation communication networks. In recent times, the detection of DDoS attacks becomes a complicated task due to high computation complexities. Several classification approaches are developed by the researchers to detect and resolve the attacks but they hardly discriminate between legitimate and attack traffics independently. In order to deal with the issue here we designed a novel DDOS detection system using the proposed weighted support vector machine kernel-based Adolescent Identity Search algorithm-Random forest (wSVMAS-RF) approach that effectively and accurately detects and classifies the malicious and benign data separately. To evaluate the proposed wSVMAS-RF approach, two modern datasets namely CICDDoS2019 and CICDoS2017 are utilized. The proposed wSVMAS-RF approach is an ensemble method that detects the DDoS attacks in transportation and application layers of the network efficiently. Before classification, the raw datasets are preprocessed using different pipelines to enhance the quality of data. In this, the dimension reduction process is executed using the Interval reduced kernel principal component analysis (IRKPCA) which helps to preserve significant data in the datasets. The proposed wSVMAS-RF approach’s efficiency is investigated by evaluating its detection performance with other state of art approaches in terms of diverse evaluation measures namely F-measure, detection rate, false positive rate, accuracy, precision and recall. The experimental results of the proposed method attained an accuracy of about 99% for the CICDoS2017 dataset and 99.74% for the CICDDoS2019 dataset.