Description

This report provides a comprehensive report that develops a taxonomy for adversarial machine learning (AML). The taxonomy includes key types of ML methods, stages of attack, attacker goals and capabilities, and attacker knowledge. AML attacks are classified as evasion, poisoning, or privacy attacks, with corresponding mitigations discussed. The report highlights open challenges in the field, such as transferability of attacks between different models, systems, and datasets. Additionally, the authors provide a glossary that defines key terms associated with the security of AI systems, aiming to assist non-expert readers.