Real-Time Phishing URL Detection Using Reinforcement Learning

Swarup Panda

Abstract: Phishing involves soliciting sensitive information by sending misleading emails that lure users to mimic legitimate websites, causing significant financial and data losses. The increase in phishing websites elevates the risk for users. Effective real-time phishing URL detection permits dynamic classification and blocking, acknowledging that malicious URLs may change over time. Accurately distinguishing between legitimate and phishing URLs constitutes a critical web-security challenge [2]. Real-time phishing URL detection methods aim to detect and dynamically block phishing URLs. Modeling real-time phishing URL detection as a reinforcement learning (RL) problem helps users avoid suspected phishing URLs because RL identifies the environment as phishing or legitimate, with the objective of dynamic classification and blocking. Reinforcement learning addresses classification as a control problem where an agent learns to make optimal decisions by interacting with an uncertain environment. Essential ingredients of an RL system include a policy, a reward signal, a value function, and an environment model. Since agents can learn value functions, policies, and models using various methods, many distinct RL algorithms have been proposed and applied to diverse problems. Leveraging diverse user interaction patterns with URLs is a promising approach for real-time phishing URL detection; if a website is widely known to be safe, it is unlikely to host malicious content.

Keywords: Serverless Security, URLs, Access Control, Phishing, Smart Contracts

How to Cite?: Swarup Panda, "Real-Time Phishing URL Detection Using Reinforcement Learning", Volume 14 Issue 8, August 2025, International Journal of Science and Research (IJSR), Pages: 317-323, https://www.ijsr.net/getabstract.php?paperid=SR25805192420, DOI: https://dx.doi.org/10.21275/SR25805192420