Powered by Restream [ Ссылка ]

This week, Apostol Vassilev from National Institute of Standards and Technology (NIST) will discuss a taxonomy of adversarial machine learning (AML) from the National Institute of Standards and Technology (NIST) Trustworthy and Responsible AI report. It explores types of attacks, attacker goals and capabilities, and mitigation methods, providing a structured understanding of AML concepts and challenges in securing AI systems. Key Takeaways: -Overview of AML attacks (evasion, data poisoning, trojans) and their impact on AI systems. -Insights into attacker objectives and knowledge levels, crucial for developing defense strategies. -Practical methods for mitigating adversarial attacks and managing AI security risks. -Identification of open challenges in AML, guiding future practices for robust AI systems.

свернуть