Performance analysis of machine learning algorithms on networks intrusion detection

Despite enormous efforts by researchers, Intrusion Detection System (IDS) still faces challenges in improving detection accuracy while reducing false alarm rates and in detecting novel intrusions. Recently, machine learning-based IDS systems are being deployed as potential solutions to detect intrusions across the network in an efficient manner. Most of them cannot perform well with large-scale or even real-time data, while the rest cannot track down evolving malicious attacks, thus putting a huge void in existing solutions. The proposed approach is an attempt to explore the possibility of developing an IDS which analyses raw network data in the form of network traffic files or server logs allowing us to simulate a real environment to accomplish testing and evaluation. Thanks to several conducted experiments, we were able to demonstrate that it is possible to improve the overall performance of learning algorithms in the field of network security by model biasing.