The Selection of Optimal Parameters for Machine Learning Methods of Detecting Malicious Requests to Web Applications

Abstract

Firewalls are still one of the key technologies for web applications protection from modern cyber threats. The in-depth protection strategy starts with isolation using firewalls and continues with other protection systems, such as intrusion detection systems. The problem with firewalls is false negatives, which can be mitigated with additional filtering tools. The use of machine learning methods is one of the possible directions in the development of defense systems. The article presents the selection of optimal parameters for several classification methods used in machine learning. For this task, a set of training data with common attacks on web applications is used.