Adversarial Training with Augmented Data: Enhancing Robustness of Machine Learning Models

Abstract

Adversarial attacks pose significant threats to the reliability and security of machine learning models, particularly in applications involving sensitive data and critical decision-making processes. Adversarial training has emerged as a promising defense mechanism to mitigate these vulnerabilities by exposing models to adversarial examples during training. However, the effectiveness of adversarial training can be further enhanced through the strategic augmentation of training data. This paper explores the integration of augmented data techniques with adversarial training to bolster the robustness of machine learning models against adversarial attacks.