Enhancing case-based reasoning approach using incremental learning model for automatic adaptation of classifiers in mobile phishing detection

Abstract

This article presents the threshold-based incremental learning model for a case-base updating approach that can support adaptive detection and incremental learning of Case-based Reasoning (CBR)-based automatic adaptable phishing detection. The CBR-based adaptive phishing detection model detects the phishing with the most suitable machine learning technique and this appropriate detection approach is endorsed by CBR technique. In such a way, the adaptive phishing detection model can address the concept drift. The threshold-based incremental learning model for a case-base updating approach will address the comprehensiveness of the knowledge in the case-base to support an incremental learning. The prototype system of our model is evaluated using nine testing feature groups of more than 20,000 phishing instances. The result shows that our adaptive phishing detection system maintains the detection accuracy while learning the new cases incrementally. The evaluation results indicate that our approach is more flexible to address the concept drift with a stable accuracy and a better performance.