Adversarial Attacks against Image Recognition Systems

Join this hands-on session to dive into the fascinating world of adversarial machine learning! We will explore how to design, evaluate, and mitigate adversarial attacks against image recognition systems using the Adversarial Robustness Toolbox (ART). ART provides a comprehensive framework to implement both white-box and black-box attacks while enabling robust defenses against such threats. Format: This interactive session will include a brief theoretical introduction followed by coding exercises. Attendees will use Python and ART to create and test adversarial examples against pre-trained image classifiers. No prior experience with adversarial machine learning is required, but familiarity with Python and basic machine learning concepts is recommended. Preparation: - Install Python 3.8 or higher. - pip install adversarial-robustness-toolbox