Analysis of Adversarial Attacks on Support Vector Machine

Abstract

This paper investigates the use of Support Vector Machines (SVMs) in sleep stage classification and their sensitivity to adversarial assaults. It illustrates the power of machine learning (ML) for precise sleep stage classification, while also emphasizing the security risks posed by adversarial attacks on ML models. Using the secML module in Python, the study investigates defense mechanisms and the robustness of SVMs against adversarial attacks. The findings highlight the significance of taking security into account when designing and deploying ML models for safety-critical applications, such as autonomous driving, cyber-security systems, healthcare, etc.