Skip to content

SafeAndSecureAI/PT-RED

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

NC / PT-RED

Brief instructions to run NC.py and pt-red.py on CIFAR-10 models.

Requirements

  • Python 3.8+
  • See requirements.txt

Install

pip install -r requirements.txt

Available model dirs

These come from the *_model.pth files in this folder:

  • badnet
  • blend
  • 1pixel
  • chess
  • clean

Ground truth target class

  • badnet: 9
  • blend: 8
  • 1pixel: 4
  • chess: 5
  • clean: none

Run

From this folder:

python NC.py --model_dir badnet
python pt-red.py --model_dir badnet

Example with another model:

python NC.py --model_dir clean

Notes:

  • The scripts download CIFAR-10 to ./data on first run.
  • pt-red.py creates ./<model_dir>/pert_estimated for outputs.

About

This is the repo for TNNLS'22 paper "Detection of Backdoors in Trained Classifiers Without Access to the Training Set" by Zhen Xiang, David J Miller, George Kesidis.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages