Hybrid Model Based Phishing Websites Detection Using Deep Learning Technique

Phishing is a type of cybercrime that poses a serious threat to online users. It involves the use of deceptive techniques to extract private information, such as login credentials and credit card numbers. Attackers often use phishing websites to obtain this data or spread harmful software. Phishing is also used as a way to identify security system weaknesses and carry out more malicious attacks, like ransomware assaults. Various methods have been employed to prevent phishing, including blacklisting, whitelisting, and machine learning (ML). Although ML techniques provide comprehensive results, they require significant computational resources due to manual feature engineering. Our approach proposes a hybrid model for detecting phishing, which utilizes deep learning techniques. We utilized the CNN-Attention-LSTM model, where CNN identifies local correlational features of URLs, and LSTM learns semantic dependencies. Our model shows promising results, achieving 97% accuracy. We also added an attention mechanism to focus on prominent features and select the most important ones, which enhances the system’s performance.