Automated Phishing Website Detection Using URL Features and Machine Learning Technique

In spite of the development of aversion strategies, phishing remains an essential risk even after the primary countermeasures and in view of receptive URL blacklisting. This strategy is insufficient because of the short lifetime of phishing websites. In order to overcome this problem, developing a real-time phishing website detection method is an effective solution. This research introduces the PrePhish algorithm which is an automated machine learning approach to analyze phishing and non-phishing URL to produce reliable result. It represents that phishing URLs typically have couple of connections between the part of the registered domain level and the path or query level URL. Using these connections URL is characterized by inter-relatedness and it estimates using features mined from attributes. These features are then used in machine learning technique to detect phishing URLs from a real dataset. The classification of phishing and non-phishing website has been implemented by finding the range value and threshold value for each attribute using decision making classification. This method is also evaluated in Matlab using three major classifiers SVM, Random Forest and Naive Bayes to find how it works on the dataset assessed.