Research objective
For decades, we have benefitted from modern cryptography to protect our sensitive data during transmission and in storage. However, we have never been able to keep the data protected while it is being processed.
In the past, cryptographic schemes that allowed processing on encrypted data were limited to partial homomorphic schemes that could support only one fundamental operation, namely either addition or multiplication but not both. Then in 2009 IBM pioneered Fully Homomorphic Encryption, which supports both fundamental operations, thus enabling the processing of data without giving access to it, however at this time it was too slow for practical use.
In recent years, thanks to algorithmic advancements, Fully Homomorphic Encryption has reached an inflection point where its performance is becoming practical. This has revolutionized security and data privacy and how we outsource computation to untrusted clouds.
Fully Homomorphic Encryption promises to disrupt major industries such as finance, healthcare, infrastructure and government by unlocking the value of data previously unreachable due to the paradox of need-to-know versus need-to-share between data custodians and data users/exploiters. For example, Fully Homomorphic Encryption makes it possible to share financial data or patient healthcare records for analytics or cross-industry collaboration without giving access to the private data.
IBM Research is closing this gap with the release of HElayers Community Edition, a Software Development Kit (SDK) for the practical and efficient execution of secure AI workloads using Fully Homomorphic Encrypted (FHE) data. To download the HELayers Community Edition Docker Container, which includes sample applications and tutorials in Jupyter Notebooks, for Windows, Linux, macOS and Linux on IBM Z mainframe please use the following link.
Publications
- A Tile Tensors Framework for Large Neural Networks on Encrypted Data
- Tile Tensors: A versatile data structure with descriptive shapes for homomorphic encryption
- Towards a Homomorphic Machine Learning Big Data Pipeline for the Financial Services Sector
- Top Brazilian Bank Pilots Privacy Encryption Quantum Computers Can’t Break
- Homomorphically Securing AI at the Edge
- Homomorphic Training of 30,000 Logistic Regression Models
- Fully homomorphic encryption method based on a bootstrappable encryption scheme, computer program and apparatus
- Efficient homomorphic encryption scheme for bilinear forms
- Efficient implementation of fully homomorphic encryption
- Fully homomorphic encryption
- Distributed computing utilizing homomorphic encryption
- Efficient two party oblivious transfer using a leveled fully homomorphic encryption