Detecto: The Phishing Website Detection

Abstract

Phishing attacks are among the most prevalent types of cybercrime that target people and businesses globally. Phishing websites mimic real websites to obtain sensitive data of users like usernames, passwords, and credit card numbers. To identify phishing websites, many people employ machine learning algorithms. These algorithms use supervised learning techniques to classify websites into the phishing or legitimate categories. Machine learning algorithms use features such as URL length, domain age, SSL certificate, and content similarity to determine whether a URL is real or fake. In recent years, authors have published papers working on the classification of websites with features by using a support vector machine and achieving 95% accuracy, and also they classify phishing websites by using a URL identification strategy or utilizing the random forest algorithm. The dataset contains a collection of URLs of 11,000+ websites. Each has 30 parameters and a class label identifying as a phishing website or not. To achieve the highest level of accuracy, we suggested a model using 32 features extracted from phishing websites and various machine learning classifiers. Every website has distinct characteristics that are categorized by trained models. We achieved 97.4% accuracy using 7 classifiers, including Naïve Bayes, logistic regression, random forest, decision tree, and gradient boosting algorithm.