
How Google is Making Private AI Practical with Homomorphic Encryption
Google introduced HEIR, an open-source compiler that enables private AI inference using homomorphic encryption. This allows computations to be performed directly on encrypted data without exposing the underlying information.
Why it matters
Industries with strict privacy laws, such as healthcare and finance, can utilize cloud AI features without risking exposure of sensitive user data.
The details
- HEIR converts pre-trained AI models to operate on encrypted inputs.
- Google partnered with hardware accelerators including Belfort, Niobium, Cornami, and Optalysys.
- Demo applications include credit card fraud detection and private content recommendations.
Show entities and relationshipsHide entities and relationships
In this article
Organizations
Technologies
Topics
Key connections
Google built and released the HEIR open-source compiler project for homomorphic encryption
Google owns Private Computing Toolkit
Google added HEIR to its Private Computing Toolkit
Google is a partner of Belfort Labs
Google partnered with Belfort Labs on hardware accelerators and private inference applications
Google is a partner of Niobium
Google partnered with Niobium on hardware accelerators for homomorphic encryption and private inference
Google is a partner of Cornami
Google partnered with Cornami to accelerate homomorphic encryption with HEIR
Google is a partner of Optalysys
Google partnered with Optalysys to accelerate homomorphic encryption with HEIR
Show 23 more connectionsShow fewer connections
Google collaborated with LG on a private deep learning recommendation model using HEIR
Google is a partner of hardshell.ai
Google collaborated with hardshell.ai to compile a credit card fraud detector using HEIR
Belfort Labs uses HEIR
Belfort Labs uses HEIR to compile private recommendation and hotword detection models
Niobium uses HEIR to compile fraud detection and network anomaly detection models
Cornami integrates hardware accelerators with the HEIR compiler
Optalysys integrates hardware accelerators with the HEIR compiler
LG collaborated on developing a private recommendation model compiled with HEIR
hardshell.ai uses HEIR
hardshell.ai collaborated on a credit card fraud detector compiled with HEIR
Georgia Institute of Technology uses HEIR
Georgia Tech researchers collaborate and conduct research using the HEIR platform
Carnegie Mellon University uses HEIR
Carnegie Mellon University researchers collaborate and conduct research on HEIR
University of California, Santa Barbara uses HEIR
UC Santa Barbara researchers collaborate and conduct research using the HEIR platform
Illinois Institute of Technology uses HEIR
Illinois Institute of Technology researchers collaborate on homomorphic encryption using HEIR
Purdue University uses HEIR
Purdue University researchers collaborate and conduct research using the HEIR platform
University of Edinburgh uses HEIR
University of Edinburgh researchers collaborate and conduct research using HEIR
Tsinghua University uses HEIR
Tsinghua University researchers collaborate and conduct research using the HEIR platform
New York University uses HEIR
New York University researchers collaborated on developing a private recommendation model compiled with HEIR
HEIR uses Homomorphic Encryption
HEIR compiles models for cryptographically-secure homomorphic encryption inference
HEIR is related to Private AI
HEIR enables practical private AI inference on encrypted data
Private Computing Toolkit uses Differential Privacy
Google's private computing technologies incorporate differential privacy
Private Computing Toolkit uses Private Information Retrieval
Google's private computing toolkit includes private information retrieval techniques
Private Computing Toolkit uses Private Set Membership
Google's privacy innovations include private set membership
Google Cloud is built with Secure Enclaves
Google Cloud provides hardware-based security solutions using secure enclaves
HEIR uses Machine Learning
HEIR converts pre-trained machine learning and AI models to run encrypted private inference on unencrypted data.
Related events
Google Showcases HEIR Open-Source Compiler for Homomorphic Encryption and Private AI
Get the weekly recap
The stories like this one, picked and explained — once a week, straight to your inbox.