Phishing Website Prediction by Using Cuckoo Search as a Feature Selection and Random Forest and BF-Tree Classifier as a Classification Method

Abstract

Phishing attacks are the threats that are ongoing in a campus and have become sophisticated as the time passes by. Target of phishing attacks are victims to affect their financial transactions and expose their personal information prone to risks. Social engineering techniques help the phishing attackers to gain and access the sensitive and personal information about a user. This may include personal details, username, password, financial details, etc. It has become very common to impersonate legitimate businesses, trick users by indulging them in business activities, and get their legitimate data as provided by them. Associative and classification algorithms can be extremely valuable in predicting phishing websites. It can give us replies about what are the most imperative cybercrime phishing site attributes and markers and how they identify with one another. WEKA tool is utilized for the usage of classifiers on an open dataset from NASA store. The inspiration driving this examination is to utilize data mining techniques and algorithms for the prediction motivation behind phishing sites and look at their viability as far as accuracy and errors. We evaluate and improve accuracy of predicting phishing websites and also reduce different types of errors, i.e., Mean Absolute Error (MAE), Relative Absolute Error (RAE), Root Relative Squared Error (RRSE), and Mean Squared Error (MSE), using cuckoo search algorithm as a feature selection technique, where random forest and BF-tree are used as classifiers.