Performance Assessment of some Phishing predictive models based on Minimal Feature corpus

Abstract

The credentials harvested normally include bank account numbers, passwords or PINs, credit card numbers, security questions, security codes etc. In most instances, vulnerability to phishing threat is due to the ease with which unsuspecting online users navigate web pages using links or URL within a body of an online message (Han et al. 2012). [...]there is an increased motivation for phishers as the number of mobile-connected devices accessing social media sites continues to grow. The limitation is often connected with superfluous training/testing time which may result in high memory overheads, delay in detection time, expensive maintenance/update etc. [...]responsiveness is used to measure prediction accuracy with commensurate processing time while the response time is used to ensure that the detection time for any window of vulnerability is reasonable and insignificant (Silva et al. 2020). In this work, we proposed an approach to examining the different state of art predictive model using reduced phishing feature corpus to resolve the uncertainties that result from performance issues (responsiveness) and other inconsistencies (response time, computational overhead etc.) in the feature set corpus.