Privacy Testing for Deep Learning

A comprehensive toolbox for model inversion attacks and defenses, which is easy to get started.

[ICML 2022 / ICLR 2024] Source code for our papers "Plug & Play Attacks: Towards Robust and Flexible Model Inversion Attacks" and "Be Careful What You Smooth For".

reveal the vulnerabilities of SplitNN

Code for "Variational Model Inversion Attacks" Wang et al., NeurIPS2021

Research into model inversion on SplitNN

📄 [Talk] OFFZONE 2022 / ODS Data Halloween 2022: Black-box attacks on ML models + with use of open-source tools

My attempt to recreate the attack described in "Model Inversion Attacks that Exploit Confidence Information and Basic Countermeasures" by Fredrikson et al. in 2015 using Tensorflow 2.9.1

a gradient-based optimisation routine for highly parameterised non-linear dynamical models

Implementation of "An Approximate Memory based Defense against Model Inversion Attacks to Neural Networks" and "MIDAS: Model Inversion Defenses Using an Approximate Memory System"