Summary
In this chapter, we covered privacy-preserving AI and the various techniques it encompasses to help us reduce the risk of exposing sensitive data to systems and attackers.
We discussed and explored a variety of techniques including data anonymization, differential privacy, distributed model training techniques such as federated and split learning, and encryption techniques such as secure multi-party computation and homomorphic encryption.
All these are part of a defense-in-depth approach to protecting sensitive data and staying compliant with privacy legislation such as the GDPR.
This completes our discussion of adversarial AI for predictive AI. In the next chapter, we will start exploring adversarial AI in generative AI.
