Abstract
In the past ten years, due to the rapid growth of the Internet, a huge number of cyber-crimes have been committed on the Internet. One of the crucial obstacles' user's encounters is the phishing website's threat, especially for login credentials and credit card information. According to the statistics, most attackers utilized PayPal (22%), Microsoft (19%), Facebook (15%), eBay (6%), and Amazon (3%) for phishing. Some phishing attempts are designed to get login credentials or infect specific people's computers. Since the stakes are so high, attackers invest a lot of effort into fooling the chosen victims. The feature selection-based Particle Swarm Optimization (PSO) method is applied in this study to detect phishing websites. Feature engineering is critical in phishing website detection systems, and detection accuracy is heavily dependent on prior knowledge of characteristics. PSO iteratively attempts to optimize a problem by improving the proposed model. Many computational methods are applied in this model to evaluate the results. The experimental findings show that the proposed PSO-based feature selection model substantially improved classification accuracy, sensitivity, specificity, f1-score, and Matthew's correlation coefficient in machine learning models. Testing the model on two prominent datasets containing phishing and legitimate website, the accuracy reaches 97.81% and 90.39% for the two datasets using the artificial neural network in both cases. Experimental findings demonstrate that the detection efficiency may be enhanced by selecting the features appropriately.