The official implementation of our AAAI 2025 paper "Detecting and Corrupting Convolution-based Unlearnable Examples", by Minghui Li*, Xianlong Wang*✉, Zhifei Yu, Shengshan Hu, Ziqi Zhou, Longling Zhang, and Leo Yu Zhang.
Convolution-based unlearnable examples (UEs) employ class-wise multiplicative convolutional noise to training samples, severely compromising model performance. This firenew type of UEs have successfully countered all defense mechanisms against UEs. The failure of such defenses can be attributed to the absence of norm constraints on convolutional noise, leading to severe blurring of image features. To address this, we first design an Edge Pixel-based Detector (EPD) to identify convolution-based UEs. Upon detection of them, we propose the first defense scheme against convolution-based UEs, COrrupting these samples via random matrix multiplication by employing bilinear INterpolation (COIN) such that disrupting the distribution of class-wise multiplicative noise. To evaluate the generalization of our proposed COIN, we newly design two convolution-based UEs called VUDA and HUDA to expand the scope of convolution-based UEs. Extensive experiments demonstrate the effectiveness of detection scheme EPD and that our defense COIN outperforms 11 state-of-the-art (SOTA) defenses, achieving a significant improvement on the CIFAR and ImageNet datasets.
| Date | Event |
|---|---|
| 2024/12/11 | The camera-ready paper is available at Detecting and Corrupting Convolution-based Unlearnable Examples! |
| 2024/12/10 | Our paper is accepted by AAAI 2025! |
| 2024/04/02 | An arXiv version including approaches to craft VUDA and HUDA unlearnable examples is updated! |
| 2023/11/30 | We have released the initial arXiv version of our paper! |
- Perform Training on Convolution-based UEs using COIN defense
python train.py --poison CUDA --arch resnet18 --coin- Perform Training on Convolution-based UEs
python train.py --poison CUDA --arch resnet18Some of our codes are built upon ECLIPSE, therefore the anconda environment is similar to ECLIPSE.
If you find COIN both interesting and helpful, please consider citing us in your research or publications:
@inproceedings{li2025coin,
title={Detecting and Corrupting Convolution-based Unlearnable Examples},
author={Li, Minghui and Wang, Xianlong and Yu, Zhifei and Hu, Shengshan and Zhou, Ziqi and Zhang, Longling and Zhang, Leo Yu},
booktitle={Proceedings of the 39th AAAI Conference on Artificial Intelligence (AAAI'25)},
year={2025}
}