Passive Captcha: Ai Driven Bot Detection For Seamless User Experience

Abstract

Traditional CAPTCHA approaches are becoming more susceptible to automated solution techniques, for which reason sophisticated and adaptive security techniques must be formulated. In this document, we present a behavior- based CAPTCHA improvement framework based on machine learning-assisted analysis of real-time user behavior. In contrast to depending on static challenges, our strategy records and assesses behavioral biometrics in the form of keystroke dynamics and mouse movement patterns in order to distinguish humans from bots.The Keystroke Detection Module regularly identifies characteristics such as dwell time, inter-key delay, and typing rhythm using a trained Random Forest classifier to detect bot-like typing behavior. At the same time, the Mouse Movement Detection Module scrutinizes trajectory characteristics such as speed, curvature, jitter, and direction changes to detect automated patterns, e.g., linear or grid movement. A majority voting system is used to make robust classification by combining multiple trajectory judgments.Both modules are combined into a real-time detection system that includes GUI support and a Flask REST API, allowing multi-modal analysis through the combination of mouse and keystroke inputs. Classification models are exportable in ONNX format to facilitate portability across platforms. To extensively verify system resilience, Selenium-based automation simulates bot behavior, including artificial typing and mouse movements. Test results indicate high precision in real-time bot detection with a seamless user experience and strong security. This research enhances CAPTCHA systems by replacing challenge-response modes with adaptive behavior-based authentication.