Detection of E-Banking Phishing Websites Using Associative Classification: A Review of Machine Learning, Feature Engineering, and Explainable Detection

Electronic banking has made financial services available at any hour, but it has also given attackers a profitable target. Phishing websites imitate genuine banking portals to collect credentials, card details and one-time passwords. Blacklists and manual checks are reactive and can miss newly created sites, so researchers have turned to machine learning, which classifies a site from features of its URL, domain and page. Many such classifiers, however, give a label without a readable reason. Associative classification combines association rule mining with classification and produces IF–THEN rules that an analyst can inspect. This paper reviews the twenty-six sources from our project synopsis that bear on this problem and groups them into six themes: traditional detection, URL and feature-based detection, machine learning, rule induction and associative classification, advanced methods, and datasets. The sources show that feature-based learning is the dominant direction and that rule-based models offer transparency. The supplied sources also leave open how associative classification compares with other classifiers, how well any model copes with unseen sites, and how recall can be kept high without raising false alarms. We outline a conceptual framework for our project; it is a proposal and has not been implemented or validated