README

This repository contains the official implementation of the methods presented in the paper Exploring Vulnerabilities of No-Reference Quality Assessment Models: A Query-Based Black-Box Method. This paper provides a query-base black-box attacking method for NR-IQA models. This repository contains the attack code for both a single image and a set of images.

Our gratitude extends to Surfree and MBS for providing the code of classification attack framework and sliency detection, respectively.

Dependencies

• python
• torch
• torchvision
• Pillow
• numPy
• scipy
• opencv-python
• h5py
• glob
• argparse

Usage

Download the files for attacked images from the LIVE dataset and other related information from Google Drive and put them into the dataset and checkpoints folders respectively.

To attack a single image, use the following command:

python attack_demo.py --incr

The option "incr" provides a strategy which increases the predicted score of the attacked image.

To attack the whole dataset, use the command in attack_live_dbcnn.sh, or use the following command:

sh attack_live_dbcnn.sh

To test the attack performance on the image set, use the following command:

python test_performance.py

Citation

@ARTICLE{blackboxIQA,
  author={Yang, Chenxi and Liu, Yujia and Li, Dingquan and Jiang, Tingting},
  journal={IEEE Transactions on Circuits and Systems for Video Technology}, 
  title={Exploring Vulnerabilities of No-Reference Image Quality Assessment Models: A Query-Based Black-Box Method}, 
  year={2024},
  doi={10.1109/TCSVT.2024.3435865}}