Hello everyone, I’m a beginner in differential privacy, I think that use Opacus is easy to implement DP-SGD, my question may be silly but I wonder if there is a way to use DP independently?
I don’t know if I’ve expressed clear enough, I mean if I add noise to dataset or output(like DP-logits) instead of gradients, is there a tool in Opacus to compute the DP part(like epsilon and delta)? If not, is there some advice to achieve this?
Thanks for your help!!
Hi @Frankozay
Thank you for your interest in Opacus.
Your question is not silly!
Opacus is a library to train a machine learning model with differential privacy with minimal code changes.
is there a tool in Opacus to compute the DP part(like epsilon and delta)
Yes, but this accounts for privacy of adding noise to the gradients, not anything else. 
I think the best is to search for tools that are for DP, and not necessarily DP in Machine Learning
Hope this helps
I got this and I’m looking forward to search other tools, thanks for your reply:)!!
1 Like