How to Boost Machine Learning Network Intrusion Detection Performance with Encoding Schemes.

Network Intrusion Detection is one of the major components of maintaining cybersecurity. This is especially crucial in Soft Targets, important places which are easily accessible, and thus more vulnerable. Real-time machine-learning-based network intrusion detection is an increasingly more relevant field of study offering important benefits to the practice of securing against cyberthreats. This paper contributes to this growing body of research by evaluating one of the problems prevailing in all machine-learning-based detectors - the notion of encoding categorical values. The use of different encoding schemes is thoroughly evaluated with the use of three different classifier types, and statistical analysis of the results is performed. The best-performing solution is proposed.