Confidential Computing and Attested Encryption

12/9/2025

This post introduces Google Cloud's Confidential Space and Oak Functions for building end-to-end confidential services. It details the challenges of trust and scalability in confidential computing and presents Oak Session as a solution for application-level, end-to-end encryption using the Noise framework. The post explains the attestation process, including JWT claims and session binding, to establish trust between the client and the confidential workload. It highlights how users can verify the integrity of the running code through reproducible builds of the Oak Functions container.

7/11/2024

Introduces the GenC open-source project, which simplifies the integration of Confidential Computing, Gemma open-source models, and mobile platforms for building private generative AI applications. Details a hybrid architecture combining on-device and cloud-hosted LLMs within Trusted Execution Environments (TEEs) for enhanced privacy and security. Explains the use of attestation for verifying TEE integrity and the GenC framework's Intermediate Representation (IR) for cross-platform deployment.

8/10/2023

Introduced new tools to expand Fully Homomorphic Encryption (FHE) capabilities for video files and other complex data. This includes a TensorFlow-to-FHE compiler for trained ML models, the Jaxite software crypto library for running FHE on hardware accelerators (GPUs, TPUs), and the Homomorphic Encryption Intermediate Representation (HEIR) open-source compiler toolchain for FHE program interoperability. The TensorFlow-to-FHE compiler can compile TensorFlow Lite models to FHE, achieving private inference in 16 seconds for a 3-layer neural network. Jaxite leverages JAX for FHE computations on accelerators. HEIR is built on MLIR to lower barriers to privacy engineering and research.