A Hybrid Edge-Client Architecture for Low-Latency Phishing Detection in Web Browsers

– Phishing remains one of the most pervasive and damaging cyber threats, deceiving users into divulging sensitive information by impersonating legitimate websites. Traditional anti-phishing solutions, primarily based on static blacklists, are increasingly ineffective against sophisticated, short-lived” zero-day” phishing attacks. This paper proposes Phishspot, a real-time antiphishing browser extension that utilizes a hybrid machine learning model to detect phishing websites. The system operates by extracting a comprehensive set of features—including URL-based, content-based, and external features—in real-time as a webpage loads. These features are then fed into a pre-trained Random Forest (RF) classifier hosted on a backend server, which returns a classification of” legitimate” or” phishing.” This hybrid architecture ensures a lightweight client-side footprint while leveraging a powerful and highly accurate server-side model. Our experimental evaluation on a large dataset of legitimate and phishing URLs demonstrates that the proposed system achieves a detection accuracy of 99.2% with a low-latency response time, making it a highly effective solution for protecting users from real-time phishing threats.