I’ve noticed that there is a feature branch implementing threshold CGGI. I was wondering what threshold scheme it is using and whether there is noise flooding in the distributed decryption? As you know, the noise flooding (or smudging) in CGGI may be too large that it moves into the message and corrupts it, as per Noah's Ark: Efficient Threshold-FHE Using Noise Flooding. How does OpenFHE mitigate this?
Thanks!
This is a prototype implementation for a work-in-progress capability. It reserves room for noise (already uses larger parameters), but does not do noise flooding yet (more analysis would need to be done). We are not actively working on this project at the moment. Threshold FHE is fully supported in OpenFHE for BGV/BFV (with proper noise flooding) and is supported for CKKS (but the configuration is trickier there as 128-bit native words are needed).