---

SeRI: Gradient-Free Sensitive Region Identification in Decision-Based Black-Box Attacks

Requirements

• Python Version: 3.11.5
• Libraries:
  • PyTorch 2.3.0
  • Torchvision 0.18.0

Installation

1. Install Python: Ensure you have Python 3.11.5 installed. If not, download it from the official Python website.

2. Install Libraries: Install the required Python libraries using the following command:

   pip install torch==2.3.0 torchvision==0.18.0

Models

Download the following pre-trained models and place them in the /code/model/ directory:

• VGG19
• ResNet50
• Inception V3
• Vision Transformer (ViT) B_32

Dataset Setup

ImageNet

Download the ImageNet dataset from the following Kaggle link:

• ImageNet Mini

Usage

To run SeRI, use the following command structure. 'norm' is the base attacker such as CGBA and ADBA; budget and budget2 are base attacker budget and SeRI budget.

python main.py --dataset=mnist-cnn --targeted=0 --norm=CGBA --zip=SeRI --budget=800 --budget2=200 --epsilon=1.0 --early=0 --beginIMG=0 --imgnum=10 --remember=1