Real-Time Mouse Data Protection Method Using GANs for Image-Based User Authentication Based on GetCursorPos() and SetCursorPos() Functions

In online services, password-based authentication, a prevalent method for user verification, is inherently vulnerable to keyboard input data attacks. To mitigate these vulnerabilities, image-based authentication methods have been introduced. However, these approaches also face significant security challenges due to the potential exposure of mouse input data. To address these threats, a protective technique that leverages the SetCursorPos() function to generate artificial mouse input data has been developed, thereby concealing genuine user inputs. Nevertheless, adversaries employing advanced machine learning techniques can distinguish between authentic and synthetic mouse data, leaving the security of mouse input data insufficiently robust. This study proposes an enhanced countermeasure utilizing Generative Adversarial Networks (GANs) to produce synthetic mouse data that closely emulate real user input. This approach effectively reduces the efficacy of machine learning-based adversarial attacks. Furthermore, to counteract real-time threats, the proposed method dynamically generates synthetic data based on historical user mouse sequences and integrates it with real-time inputs. Experimental evaluations demonstrate that the proposed method reduces the classification accuracy of mouse input data by adversaries to approximately 62%, thereby validating its efficacy in strengthening the security of mouse data.