Attacks aimed at reconstructing or cloning proprietary machine learning models through targeted queries to the API or exposed service. Includes model stealing techniques, parameter extraction, functional reverse engineering, and strategies for exfiltrating predictive logic from ML-as-a-Service systems.