Intrusion Detection of Imbalanced Network Traffic

Abstract

With the growth of the Internet, cyber-attacks are evolving at a rapid pace, and the cyber security scenario is bleak. Machine and Deep Learning approaches for intrusion detection network analysis, as well as a brief instructional overview of each Machine Learning/Deep Learning method. Using temporal or temperature correlations, those presenting each approach were summarized. They cover some of the most often used datasets in Machine learning, highlight the limitations of employing these algorithms for cyber security, and offer future opportunities because data is so crucial in these methodologies. There is a lot of effort being done to enhance intrusion detection algorithms, but study into the information to train & test the model is important since higher data quality can improve offline intrusion detection. The study examines the KDD data set in terms of four classes: Basic, Content, Traffic, and Host, with all data characteristics being classified using MODIFIED RANDOM FOREST (MRF). For Intrusion Detection System (IDS), the analysis is done using two popular assessments metrics: Detection Rate (DR) and False Alarm Rate (FAR). The influence of each of four classes of qualities on DR and FAR is displayed as per the output based on data.