Enhancing Web Security: Hybrid Deep Learning Model for Robust URL Phishing Detection

Security is the primary difficulty in the realm of the Internet and different kinds of electronic communication. It’s significant to remember that phishing is only one of many methods that can be used to compromise a user’s security on the internet; in fact, it’s become one of the greatest and most successful cyber dangers, responsible for losses of hundreds of millions of dollars & millions of data breaches annually. To steal sensitive user information, phishers create websites that seem almost identical to legitimate ones. Users’ private data might be compromised, putting them at risk of identity theft or financial loss. To identify phishing sites, current anti-phishing methods rely on experts manually extracting features from malicious sites and external services. In particular, it takes a lot of effort and skill to extract phishing features using these methods, which is a drawback. In addition, phishing site detection is slowed when external services are used. That’s why it’s crucial to have a method that can effectively identify phishing websites. To identify malicious websites, this research develops a Deep Learning (DL) model that combines conventional CNNs with BiLSTMs. No content or external services are required for this approach to accurately forecast whether or not a URL is legitimate. The proposed CNN-BiLSTM model is accurate in experiments, with an accuracy of 98.8% being achieved. The results of the DL-based hybrid system prove the superiority of the suggested phishing detection strategy.