Peter Kairouz comments

Results 8 comments of


                                            Peter Kairouz

Create abstract DPAverageQuery

TF Privacy is designed to be compatible with TFF. Integrating these two should be easy, and we are going to do it very soon.

Calculating the privacy budget in the absence of additive noise

Hello, Right now, we only support that subsampled Gaussian mechanism. Can you be a bit more specific about what other "adjustments on the queries" you have in mind? Thanks, peter

Calculating the privacy budget in the absence of additive noise

Unfortunately, clipping the gradients alone does not provide differential privacy. You need to add noise, in addition to clipping, to be able to obtain differential privacy. On Mon, Nov 2,...

Calculating the privacy budget in the absence of additive noise

Adding noise alone also does not provide DP. You need to clip the gradients to bound the sensitivity of the query you are computing. The clipped gradients are then aggregated...

Calculating the privacy budget in the absence of additive noise

We currently do l2 clipping (i.e. we make sure that each gradient has a bounded l2 norm). As mentioned before, this translates into a bound on the l2 sensitivity of...

[differential_privacy] Learning rates used for Adaptive Clipping experiments

Using a server learning rate = 1 should be just as good as anything else. I believe that's what we observed.

[differential_privacy] Learning rates used for Adaptive Clipping experiments

My impression is that yes, this should work well for all noise multipliers and clipping quantiles. I will check with the paper authors and get back to you on this...

[distributed_dp] Including package versions into the requirements file

Could you please expand more on where exactly you are doing? Are you creating a custom iterative process or using one that we are providing in the repo? Could you...