An organization's AI risk team wants to assess the attack surface of an AI system before deployment. Which framework provides the MOST comprehensive vocabulary for cataloguing adversarial threats specific to AI and ML systems?
Select an answer to reveal the explanation.
Short Explanation and Infographic
Here's the deal — b is correct because MITRE ATLAS is specifically designed to catalogue real-world adversarial attack techniques targeting AI and ML systems, organized in a structure analogous to MITRE ATT&CK. It covers tactics such as ML model access, training data poisoning, inference API access, and model evasion.
Full explanation below image
Full Explanation
B is correct because MITRE ATLAS is specifically designed to catalogue real-world adversarial attack techniques targeting AI and ML systems, organized in a structure analogous to MITRE ATT&CK. It covers tactics such as ML model access, training data poisoning, inference API access, and model evasion. OWASP Top 10 (A) focuses on web application vulnerabilities rather than ML-specific adversarial techniques. NIST CSF 2.0 (C) is a general cybersecurity framework not specific to AI attack taxonomies. ISO/IEC 27005 (D) is a general information security risk management standard.