Publication

Abstract : Lazy classifiers are the local learner, where the model is built only when the test instances are present at hand. In this paper, five lazy learners are explored using ISCXIDS2012 and CICNTTor2016 datasets provided by the Canadian Institute of Cybersecurity. In order to explore these classifiers, 11 performance measures are considered for each dataset separately. The result shows that KStar classifier suitable for ISCXIDS2012 and both IB1 and IBkLG classifiers works well in the CICNTTor2017 dataset. Therefore, these classifiers can be used as an adaptive intrusion detection engine to detect possible intrusions by preserving inherent late learning property of lazy classifiers.