Automated AI System for Online Phishing Detection and Mitigation

This paper presents the critical challenge of phishing attacks by using advanced machine learning (ML) approaches to efficiently identify and mitigate malicious websites. Phishing websites betray users by impersonating legitimate entities to extract sensitive information. This paper emphases on developing and implementing a robust ML model capable of differentiating between legitimate and phishing websites through a comprehensive analysis of URL features, website content, and associated metadata. The proposed approach employs a diverse dataset that includes both phishing and legitimate websites, with features extracted using both traditional approaches and innovative methods. Various ML approaches are explored, such as Decision Trees, Support Vector Machines, Random Forest (RF), and Neural Networks. These approaches are evaluated for accuracy, recall, precision, and F1-score to determine the most effective model. Feature selection approaches are employed to discover the most significant indicators of phishing activity, improving the model's performance and interpretability. The RF Classifier demonstrated superior performance with an accuracy of 96.6%.