DETECTING BOTS IN WEB APPLICATIONS BASED ON USER BEHAVIOR USING MACHINE LEARNING

Abstract

This article proposes an innovative machine learning (ML)-based approach to distinguish between users and bots in web applications. Due to the declining effectiveness and user inconvenience of traditional methods (e.g., CAPTCHA), the authors suggest detecting bots by analyzing interactive user behavior, such as mouse movements, keyboard inputs, scrolling patterns, and other parameters.

The study employs a lightweight neural network model in ONNX format, demonstrating high accuracy (93.5%) and low error rates in real-time bot detection. Key advantages of the model include minimal impact on user experience, fast processing (<50ms), and adaptability to various devices