Abstract

Phishing attacks remain a significant threat to online security, as cyber criminals regularly create new malicious URLs to trick or deceive users. Traditional detection methods such as blacklist-based detection, rule based detection, and signature-based detection are not useful for real time detection. All of these methods are slow and do not give high accuracy. So, this research paper proposes a hybrid method to obtain high accuracy and real time detection. Here we use Random Forest and DistilBERT together for better performance. In the first step, we used Random Forest to perform fast filtering using URL features. If the URL is suspicious, then it goes to the second stage, where we use DistilBERT to check for additional contextual and semantic patterns. This hybrid model performs better than individual models in terms of accuracy, precision, recall, F1 score, and also in real time phishing detection. It achieves 96.60 percent accuracy.