Detection of phishing websites via machine learning algorithms and application

Phishing is one of the most popular types of attacks that have recently created a security risk for internet users. The web sites used for this offensive method can easily be multiplied and make it difficult to determine the social engineering methods used during the attack. Using the classification algorithms, the attacks of the web sites prepared for this purpose can be detected instantaneously. In this study, models were trained on URL sets obtained from PhishTank and Google search results using Logistic Regression, KNN and Naive Bayes algorithms. The success of the models; Accuracy, Sensitivity, Recall, and F-score were evaluated and the successes of the algorithms were compared. The Logistic Regression algorithm is found to be that more successful than the other algorithms.