add test to show that IPA fails for cap 1, 2, 3, and 4 #1346
Conversation
It is still gross that we use 3 bit trigger values for everything, but oh well...
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #1346 +/- ##
==========================================
+ Coverage 93.44% 93.48% +0.04%
==========================================
Files 209 211 +2
Lines 34512 34959 +447
==========================================
+ Hits 32249 32683 +434
- Misses 2263 2276 +13 ☔ View full report in Codecov by Sentry. |
| @@ -147,13 +147,15 @@ where | |||
| #[cfg(not(feature = "relaxed-dp"))] | |||
| let padding_params = PaddingParameters::default(); | |||
| match config.per_user_credit_cap { | |||
| 1 => oprf_ipa::<_, BA8, BA3, HV, BA20, 1, 256>(ctx, input, aws, dp_params, padding_params).await, | |||
There was a problem hiding this comment.
Using BA3 for trigger values with cap 1 is gross, but the input is assumed to be fixed size, so it is a much bigger change to make if we want it
| 8 => oprf_ipa::<_, BA8, BA3, HV, BA20, 3, 256>(ctx, input, aws, dp_params, padding_params).await, | ||
| 16 => oprf_ipa::<_, BA8, BA3, HV, BA20, 4, 256>(ctx, input, aws, dp_params, padding_params).await, | ||
| 32 => oprf_ipa::<_, BA8, BA3, HV, BA20, 5, 256>(ctx, input, aws, dp_params, padding_params).await, | ||
| 64 => oprf_ipa::<_, BA8, BA3, HV, BA20, 6, 256>(ctx, input, aws, dp_params, padding_params).await, | ||
| 128 => oprf_ipa::<_, BA8, BA3, HV, BA20, 7, 256>(ctx, input, aws, dp_params, padding_params).await, | ||
| _ => panic!( | ||
| "Invalid value specified for per-user cap: {:?}. Must be one of 8, 16, 32, 64, or 128.", | ||
| "Invalid value specified for per-user cap: {:?}. Must be one of 1, 2, 4, 8, 16, 32, 64, or 128.", |
There was a problem hiding this comment.
I don't know why we picked cap sizes to be a power of two, but I tried to stick with the same approach.
@eriktaubeneck, @shinta-liem would that be sufficient for in-market test to support cap of 1, 2 and 4?
There was a problem hiding this comment.
yes, that's plenty sufficient
@akoshelev here is a few basic tests to show the that IPA doesn't work for per user credit cap of 1, 2, 3, or 4.