Protecting Patient Data

The SAIL platform protects patient data using end-to-end encryption and by performing all computations on the data in a Secure Computing Node running in a Trusted Execution Environment, a hardware-enforced confidential computing environment. Our Digital Contract technology ensures that only the holder of a valid digital contract may trigger computations on the data set. Arbitrary code can not be run against the data, further securing the patient data against malicious actors and inadvertent data leaks. By using the same technology in the inference stage in production systems, the AI model can run on data “blindly”; that is, the model owner and the party hosting the model can be shielded from seeing the data inputs or the prediction outputs from the model.

Protecting Commercial IP

Most cloud-based or data-owner-hosted analytics platforms have no technical guarantee of confidentiality of the commercial intellectual property of the data consumer via exposed queries. As an example, a biotech company may want to query a patient genomic database without exposing the genes of interest to the data owner or the analytics system provider. The SAIL platform guarantees confidentiality by encrypting the queries end-to-end, and also running all computations in a Trusted Execution Environment. Neither SAIL nor the data owners have access to the queries and computations.

SAFE Functions

SAFE (Safe Algorithms for Federated Enclaves) Functions are vetted computational algorithms that operate on data confidentially within the Secure Computation Node. Because these functions are run within the Trusted Execution Environment, the data themselves are never visible to the outside world. By limiting the users to use only these SAFE Functions, the system drastically reduces the possibility of “malevolent algorithms,” which are often the Achilles’ heel in Federated Learning set ups.

Our Work

Kidney Cancer Association

The KCA will leverage SAIL’s platform to make use of critical medical data from hospitals around the country while maintaining patient privacy.
The resulting research has the potential for unprecedented breakthroughs in kidney cancer treatment.

View more

Questions about Federated Learning?