SourceHow Google is Making Private AI Practical with Homomorphic Encryption
Industries with strict privacy laws, such as healthcare and finance, can utilize cloud AI features without risking exposure of sensitive user data.
- Google introduced HEIR (Homomorphic Encryption Intermediate Representation) to enable cryptographically-secure private AI inference on encrypted data without revealing underlying user information
- HEIR converts pre-trained AI models operating on unencrypted data into models operating on encrypted inputs, aiming to provide a one-click encrypted inference solution for production applications
- Google partnered with homomorphic encryption hardware accelerator companies including Belfort, Niobium, Cornami, and Optalysys
- Google demonstrated four private inference applications compiled with HEIR: a deep learning recommendation model, credit card fraud detection, network traffic anomaly detection, and a hotword detector
- HEIR has supported collaborations across academic institutions including Georgia Tech, Carnegie Mellon, UC Santa Barbara, Illinois Institute of Technology, Purdue, University of Edinburgh, Tsinghua University, and New York University