Ensemble-based Semi-supervised Learning Approach for DoS Detection Using Feature Selection Algorithm

Abstract

Interconnected systems such as database servers, web-servers are now under threats from network attackers. A single attack on a computer or network system can lead to significant damage. Denial of Service (DoS) is a severe form of network attack that is used against an information system in order to block legitimate users from accessing the infected system. In a DoS attack, the attacker usually generates enormous packets by a large number of compromised computers and can easily force victims out of service within a short period of time. Therefore, an ensemble semi-supervised learning approach is proposed in this paper using different classification techniques. These techniques are Random Forest, Decision Tree and Extreme Gradient Boosting. To evaluate the detection performance of the proposed approach, extensive experiments have been conducted on the benchmark datasets such as KDD'99, NSL-KDD, and UNSW-NB15. Besides, the proposed approach has been compared with other machine learning techniques to validate based on the same datasets.