Advanced Phishing Detection System Using Federated Learning

Jha A and Raj A

Published on: 2025-12-31

Abstract

Phishing attacks continue to pose a serious risk to individuals and organizations worldwide. A major limitation of centralized phishing detection systems is the challenge of preserving data privacy, along with increasing complexity as new attack techniques emerge. This work proposes an enhanced phishing detection method that leverages federated learning, where user datasets are processed locally to maintain privacy while achieving improved performance. The method integrates multiple local models from user devices into a global model, which enhances detection accuracy compared to standalone models and ensures raw data remains secure. Experimental evaluation across diverse datasets demonstrates that the proposed system achieves higher scalability, stronger adaptability to evolving threats, and better protection of sensitive user information.