A novel technique for converting nominal attributes to numeric attributes for intrusion detection

Abstract

Intrusion Detection has been a popular area of research due to increase in number of attacks. Intrusion Detection is a classification problem, in which some of the attributes are nominal. Classification algorithms like Support Vector Machine, Extreme Learning Machine, Neural Network etc. are not capable of handling nominal features. This leads to the need of method for converting nominal features to numeric features. None of research articles published till date have evaluated the appropriate method of nominal to numeric conversion for intrusion detection problem. This work explores Target Methods, Dummy Methods and Influence Value Method for Intrusion Detection to convert nominal attributes to numeric attributes. This work also proposes a new method for nominal to numeric conversion, which performs better than existing methods. The results presented in this paper evaluated using NSL-KDD.