PhishNet: Smart Website Security with Conversational AI

Phishing remains a significant cybersecurity threat, enabling adversaries to impersonate legitimate digital platforms and compromise sensitive user information. Traditional blacklist-based detection mechanisms are inherently reactive and often fail to identify newly generated phishing domains in real time. This study presents PhishNet, a hybrid phishing detection framework integrating blacklist intelligence, heuristic URL analysis, and conversational AI within a browser-based architecture. The proposed system employs structural and lexical feature evaluation combined with external threat verification to classify URLs as safe, suspicious, or malicious. Experimental evaluation on a dataset of 6,000 labeled URLs achieved 97.8% accuracy, 97.1% precision, and 98.2% recall, with an average latency of 265 ms. The results demonstrate that the hybrid approach effectively balances detection accuracy, computational efficiency, and user interpretability.