Mitigating Cybersecurity Threats in ML-Based Systems (CTA-2114)

Machine Learning (ML)-based systems have an intrinsic set of challenges, including cybersecurity, that need to be addressed during their development and specifically monitored during their deployment and use. ML-based systems are not immune to cybersecurity breaches; additionally, they may be subject to adversarial attacks on their models, algorithms, or data.

This document identifies methods for mitigating cybersecurity threats to and privacy concerns in ML-based systems by addressing the unique considerations of ML-related products.