Using Machine Learning for Network Intrusion Detection
Divya Gupta, Ashendra Kumar Saxena · 2022
To prevent cyberattacks from becoming catastrophic, anomaly detection is crucial since it may warn managers to potentially malicious network activities. There have been significant developments, but the vast majority of available commercial Intrusion Detection Systems still depend on signatures to detect hackers. The need to regularly update the signature database when new attack pattern signatures become available is a fundamental limitation of signature-based techniques, making them inappropriate for accurate network anomaly detection. Recently, machine learning-based classification techniques have gained traction in the anomaly detection community. We apply seven different machine learning techniques to the Kyoto 2006+ dataset and calculate the information entropy to compare their performance. Overall, most machine learning approaches were shown to have 90% or accurate results, recall, and accuracy on this dataset. However, when comparing the seven algorithms in terms of the area under the Receiver Operating Curve, Radial Basis Function comes out on top.