Limitations and Roadmap
Support for multiple cloud platforms
The first reference implementation is based on Azure (using the serverless Confidential Containers offering). In due course of time, we will release reference implementations for GCP (using Confidential Spaces) and AWS (using Nitro Enclaves).
The reference implementation supports a pipeline with basic aggregation operation and training. We will extend the implementation to support richer class of operations e.g., data transformations, data augmentations without compromising privacy.
We will provide TDPs with better tools e.g., for managing their identities, datasets, and privacy budgets.