privML / privacy-evaluator

The privML Privacy Evaluator is a tool that assesses ML model's levels of privacy by running different attacks on it.
MIT License
17 stars 17 forks source link

Add description about how to optimize attack model. #229

Closed marisanest closed 3 years ago

marisanest commented 3 years ago

This is just a draft. What is missing are some descriptions in the notebook about how to optimize attack models. And I am also not sure about the text I added to the README. @Friedrich-Mueller what do you think? I added now some sort of code explanation to the README, which might be not a good place to do so, but it seemed reasonable. Might be good for the user to have some sort of explanation of our framework along with more general, scientific info about the attacks.

Closes #195.

Friedrich-Mueller commented 3 years ago

Thanks, ill look into it some time today Not sure if I will be able to before our meeting

marisanest commented 3 years ago

Do not merge yet, I'm waiting till all other pull requests with notebook changes involved are done so I can add my notebook changes as well.

marisanest commented 3 years ago

Can now be merged if reviews are approved :+1: