Multi-Class SVM & Random Forest Based Intrusion Detection Using UNSW-NB15 Dataset
A M Nacif Jose, Avirup Mukherjee, Joydeep Saha, Anirban Bhaumik, Kamalesh Datta, Radhashyam Patra, Aradhana Behura · 2024
Intrusion Detection or malicious node detection (IDS) show a important role in safeguarding computer networks system from security threats. However, traditional IDS methods often struggle with high false positive rates and limited detection accuracy. This paper presents an advanced IDS that harnesses the power of a multi-class support vector machine (SVM) and integrates random forest technique for feature selection, specifically targeting the UNSW-NB15 dataset. The motivation behind this research is to enhance IDS performance by leveraging robust machine learning techniques. The research problem centers on improving the accuracy and efficiency of detecting various network attacks while reducing false positives. Our approach involves meticulously preprocessing network traffic data to extract relevant features, utilizing random forest to identify the most informative ones, and then employing these features to train a multi-class SVM classifier. This classifier effectively categorizes network traffic into different attack types. The results demonstrate a significant improvement in prediction accuracy and a decrease in rate of false positives, thereby enhancing the overall effectiveness of IDS in protecting computer networks against security threats.