Detection of fake web pages and phishing attacks with rabbit optimization algorithm

Phishing attacks are a type of deception-based attack in which the thief directs Internet users to fake sites. Fake sites look very similar to real sites, so users easily trust them and reveal their valuable information, such as usernames and passwords. A practical way to detect phishing attacks is to use machine learning techniques. In order to reduce the detection error of phishing attacks, a two-step approach is presented in this paper. The first stage selects the essential features using the rabbit optimization algorithm. This feature selection technique identifies the most relevant features for the detection of phishing attacks and is delivered to three classifications, MLP, RF, and XGBoost, to decrease the detection error of phishing attacks in embedded learning with voting. The tests performed in the MATLAB environment are performed on two datasets: UCI and Tan. Experiments show that the proposed method in the UCI dataset has accuracy, sensitivity, and precision of 97.82%, 97.51%, and 97.62%, respectively. If the Tan dataset is used, then the accuracy, sensitivity (recall), and precision of the proposed method in detecting phishing attacks are 97.91%, 97.88%, and 97.52%, respectively. The results show the proposed method’s superior performance in detecting phishing attacks. The proposed approach to detecting phishing is more accurate than the RF, MLP, XGBoost, BiLSTM, and RNN methods.