PhishGuard: Integrating Fine-Tuned Large Language Models (LLMs) into Password Management

Abstract

In the digital age, phishing attacks have been a persistent security threat leveraged by traditional password management systems that are not able to verify the authenticity of websites. This paper presents an approach to embedding sophisticated phishing detection within a password manager’s framework, called PhishGuard. PhishGuard uses a Large Language Model (LLM), specifically a fine-tuned BERT algorithm that works in real time, where URLs fed by the user in the credentials are analyzed and authenticated. This approach enhances user security with its provision of real-time protection from phishing attempts. Through rigorous testing, this paper illustrates how PhishGuard has scored well in tests that measure accuracy, precision, recall, and false positive rates.