search

privacytrustlab / ml_privacy_meter

Privacy Meter: An open-source library to audit data privacy in statistical and machine learning algorithms.
MIT License
556 stars 99 forks source link

Old tutorials with restructured code #81

Closed nitincic closed 1 year ago

nitincic commented 1 year ago

There are some tutorial in the archive folder from the earlier version of the library which work for whitebox attacks (exposing gradients etc). I don't see similar options with the restructured code. Is it still possible to audit those attacks or have those attacks been renamed?

changhongyan123 commented 1 year ago

Hi @nitincic , thank you for your question. The restructured code does not provide examples of the white-box attacks. However, you can specify how to get the gradient information (e.g., PyTorch) in the model.py and define your own Metric, which infers members using the ModelGradient. For example, you can replace ModelLoss in audit.py. We will provide white-box attack examples using restructured code soon.