Question:During a security audit, analysts discover that an attacker repeatedly queried a black-box ML model...

ISO/IEC 23894:2023 (Artificial Intelligence Risk Management) and NIST SP 800-207A define Membership Inference Attacks (MIA) as:

“An adversary attempts to determine whether specific data was used in the training phase of a machine learning model.”

This is a privacy threat and can lead to data breaches, especially with personally identifiable information (PII). It differs from data poisoning, which manipulates the training process, and backdoors, which alter behavior intentionally.