Attribution of Gradient Based Adversarial Attacks for Reverse Engineering of Deceptions

Code to replicate results in this paper: https://arxiv.org/pdf/2103.11002.pdf

1. Create a directory called "data/" in the root of this repo

2. Download the dataset here

3. Unzip into the "data/" directory

4. Create a virtualenv, and install the packages in requirements.txt

5. Create the adversarial samples by executing run_attacks.py, in src/data/

6. Train each of the 3 models in the paper, by running the following in src/models/

   python train.py co_occur GPU_ID

   python train.py laplace GPU_ID

   python train.py direct GPU_ID

7. Produce test results by running this for all 3 methods:

   python predict.py {co_occur,laplace,direct} GPU_ID

8. Print out results and confusion matrices as arrays and latex code:

   python eval.py {co_occur,laplace,direct}