A Comprehensive Analysis of Detection Methods for Image-Scaling Attacks

This repository belongs to the publication:

---

Erwin Quiring, Andreas Müller and Konrad Rieck. On the Detection of Image-Scaling Attacks in Machine Learning. Proc. of Annual Computer Security Applications Conference (ACSAC), 2023.

---

The Bibtex key:

@inproceedings{QuiMueRie23,
        title={On the Detection of Image-Scaling Attacks in Machine Learning},
        author={Erwin Quiring and Andreas Müller and Konrad Rieck},
        year={2023},
        booktitle={Proc. of Annual Computer Security Applications Conference ({ACSAC})},
  }

Background

In short, image-scaling attacks enable an adversary to manipulate images, such that they change their appearance/content after downscaling. In particular, the attack generates an image A by slightly perturbing the source image S, such that its scaled version D matches a target image T. This process is illustrated in the figure above.

In this paper, we study different detection methods against image-scaling attacks.

For a root-cause understanding and prevention techniques, please look at this paper from Quiring et al..

Getting Started

First, make sure you have Python and Anaconda. In the following, we assume you are using Anaconda (any other system like virtualenv would also work).

Next, get our repository (that builds up on the repository from Quiring et al.):

git clone https://github.com/EQuiw/2023-detection-scalingattacks.git
cd 2023-detection-scalingattacks/scaleatt

Create a python environment:

conda create --name scaling-attack python=3.7
conda activate scaling-attack

Install python packages and compile cython extensions:

pip install -r requirements.txt
pip install -r requirements_jupyter.txt
python setup.py build_ext --inplace

Only in case of problems: check the SETUP-README in the scaleatt directory for a detailed introduction how to set up the project.

Tutorial

Now, in the scaleatt directory, just start Jupyter in the console.

jupyter notebook

1. You will be redirected to your web browser where you can run the notebooks.
2. Open Tutorial-Detection.ipynb
3. This file explains the attack, detection methods, and adaptive attacks step by step.