Michael Shoemate

This issue is because the SDK has a dependency on OpenDP 0.4. OpenDP 0.5 supports M1, so at this point we just need an update to the SDK to support...

These cites will naturally be embedded in the proofs. This will get resolved as the proofs mature. EDIT: This is a citation for the opendp repo, not the constructors within....

I think the file structure here should get the same treatment as the randomized response proof PR-- but it will need to be split up. There's a section on domains,...

Another useful resource is this comment, explaining how to use the existing C api exposed by the rust library. https://github.com/opendp/opendp/issues/235#issuecomment-902855244

I rebased this October 2021 PR today. Didn't encounter any conflicting changes, aside from the crate merge (which is a good thing, indicating that these interfaces are pretty stable).

This algorithm above to batch graphs has been implemented, and the privacy composition is exposed in a function. At this point the function just returns linear composition. Essentially, library integration...

Thank you for looking into, and opening this! Understood, the result has the precision of the least precise summand, which is data-dependent. I'm thinking about mitigations. The only way MPFR...

Hi @TedTed, the newsletter is referring to changes made in 0.4, our last release from December. I did do hardening related to rounding of floating-point types in that release. I...

@TedTed I've become skeptical of the integerization approach for approximating laplace noise. In the limit as granularity approaches zero, the granularity becomes smaller than the ulp of the input argument....

I don't currently have plans to add gRPC service definitions, but I'm not opposed to adding them. We don't currently use gRPC because, for my use case, gRPC hasn't been...