Selection of robust feature subsets for phish webpage prediction using maximum relevance and minimum redundancy criterion
Phishers usually evolve their web exploits to defeat current anti-phishing community. Accordingly, that becomes a serious web threat and puts both users and enterprises at the risks of identity theft and monetary losses day by day. In the literature, most computational efforts were dedicated to just...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Published: |
Asian Research Publishing Network (ARPN)
2015
|
| Subjects: | |
| Online Access: | http://eprints.utm.my/id/eprint/55407/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Phishers usually evolve their web exploits to defeat current anti-phishing community. Accordingly, that becomes a serious web threat and puts both users and enterprises at the risks of identity theft and monetary losses day by day. In the literature, most computational efforts were dedicated to justify well-performed phishing detection against evolving phish exploits. However, facets like exploration of new and predictive features, selecting minimal and robust features compactness still raise as key challenges to optimize the detection scenarios over vast and strongly interrelated web. In this study, we proposed a set of new hybrid features, and refine it as few, maximum relevant, minimum redundant, and robust features as possible. In the presence of a machine learning classifier and some assessment criteria that recommended for this purpose, the reported results experimentally demonstrated that our remedial scenario could be used to optimize a phish detection model for any anti-phishing scheme in the future. |
|---|