Evaluation of Intelligent Intrusion Detection Models

Abstract

This paper discusses an evaluation methodology that can be used to assess the performance of intelligent techniques at detecting, as well as predicting, unauthorised activities in networks. The effectiveness and the performance of any developed intrusion detection model will be determined by means of evaluation and validation. The evaluation and the learning prediction performance for this task will be discussed, together with a description of validation procedures. The performance of developed detection models that incorporate intelligent elements can be evaluated using well known standard methods, such as matrix confusion, ROC curves and Lift charts. In this paper these methods, as well as other useful evaluation approaches, are discussed.