-
Currently, `MeasurementSpec` contains a `SketchConfig`. This is a very flexible configuration used by the `AnySketch` library. The CMMS API need not expose this level of flexibility for Measurement cr…
-
Hi, I can't tell if this is support, or a feature request; please direct me elsewhere if needed.
How can I use a `PrivacyEngine` *without* gradient clipping (i.e., only add noise)?
I am not conc…
-
Implement support for policies that avoid side-channel leaks from the response. The policies can be specified using protobuf as suggested by @tiziano88:
```protobuf
message Policy {
// A fixed …
-
## Question
I want to using some local differential privacy mechanisms during the training process, how can I do that?
-
From @RomainGoussault:
> [Moving beyond "algorithmic bias is a data problem"](https://www.cell.com/patterns/fulltext/S2666-3899(21)00061-1)
Question it raises: should we add an evaluation element …
bowni updated
3 years ago
-
Hi developers,
- I read a few suggestions about computing epsilon for federated learning in differential Privacy but still get some confusions. Therefore I hope you can help me make clear some points…
-
Optionally, Prio servers should add Laplace-distributed noise to their shares of the final statistic before publishing them. In this way, the final output of the system can be differentially private w…
-
Hi, I am running Federated Learning with differential privacy folder.
It seems the cifar10 is only ok with 10 clients (number of client = number of class). When changing the number of client differen…
-
I'm trying to have superFreq output to an absolute path rather than using the base path where the bam files for the quality contorl samples are. The line below seems to force creation of the R and plo…
-
Our current design tightly couples information that we need for differential privacy (such as NameCompare) with engine-proprietary connection information that is used to establish a database connectio…