A Literature Review On Machine Learning-Based Phishing Detection Systems

Abstract

This paper presents Threat Scout, a client-side hybrid framework for real-time phishing URL detection. The system integrates machine learning models with heuristic analysis to identify malicious websites that attempt to steal sensitive user information. Unlike traditional blacklist-based approaches, Threat Scout operates offline within the browser, ensuring privacy and low latency. To improve robustness, the system combines lexical, domain-based, and content features with adversarial defense techniques such as Document Object Model (DOM) structure analysis and visual similarity checks. By delivering immediate, context-aware alerts through a lightweight browser extension, Threat Scout empowers users with proactive protection against phishing attacks. The framework is designed to be scalable, resource-efficient, and user friendly, enabling deployment across multiple browsers. This paper details the architecture, methodology, and expected impact of Threat Scout in strengthening client-side web security.