## **PoC** [The code](https://github.com/ffhibnese/Model-Inversion-Attack-ToolBox) Run attack algorithms with scripts in `./attack_scripts/`. Use the command below, and results will be saved to `./results/<ATTACK_METHOD>/` by default: `python attackscripts/<ATTACKMETHOD>.py` The framework also contains a defensive toolkit - which can be found [here]() ## **Details** By abusing access to the pre-trained models via white or black box access, this toolkit will allow the operator to invert a model,  reconstructing high-fidelity data that closely aligns with the private training data. 🚀 MIA repository offers an open-source Python test suite for conducting model inversion attacks. It features unified implementations of advanced model inversion methods for easy and fair comparison. [Paper](https://arxiv.org/abs/2402.04013) ### ATT&CK Matrix