Anomaly-Based Network Intrusion Detection System: A Machine Learning Approach

Abstract

At the present time, anomaly detection has attracted the attention of many re- searchers to overcome the weakness of signature-based IDSs in detecting novel attacks, and NSL-KDD benchmark data set is the most used in the literature which it was generated a decade ago, therefore, it does not re ect modern network tra c and low footprint attacks. A new data set has been devel- oped named UNSW-NB15 network data set that came to solve the issues of NSL-KDD in which it has been used in this project to evaluate Anomaly-based NIDs by using di erent machine learning methods such as Logistic Regression, Decision Tree, Random Forest and particularly an Arti cial Neural Network with both binary and multi-class classi cation that we have discussed their re- sults in details and they have been compared with some previous related works. keywords- Network Securtiy; NIDS; UNSW-NB15 data set; Machine learn- ing; Multi Layer Perceptron.