albblgb/Deep-Steganalysis

PyTorch implementation of the classical image steganalysis networks, compatible with both grayscale and color images, and supporting multiple image sizes.

Deep-Steganalysis

The reproduced literature is listed as follows:

• XuNet (SPL2016): Structural Design of Convolutional Neural Networks for Steganalysis.
• YeNet (TIFS2017): Deep Learning Hierarchical Representations for Image Steganalysis.
• StegNet (IH&MMSec2017): Fast and Effective Global Covariance Pooling Network for Image Steganalysis.
• SRNet (TIFS2019): Deep Residual Network for Steganalysis of Digital Images.
• ZhuNet (TIFS2020): Depth-Wise Separable Convolutions and Multi-Level Pooling for an Efficient Spatial CNN-Based Steganalysis.
• SiaStegNet (TIFS2021): A Siamese CNN for Image Steganalysis.

Dependencies and Installation

• Python 3.8.13, PyTorch = 1.11.0

• Run the following commands in your terminal:

  conda env create -f env.yaml

  conda activate pyt_env

Get Started

Training

1. Change the code in config.py

   line4: mode = 'train'

   line17: train_data_dir = ''

   line18: val_data_dir = ''

   line20: stego_img_height =

   line21: stego_img_channel =

2. Run python *net.py. For example, python srnet.py

Testing

1. Change the code in config.py

   line4: mode = 'test'

   line19: test_data_dir = ''

   line36-41: pre_trained_*net_path = ''

2. Run python *net.py

• The trained steganalysis networks will be saved in 'checkpoint/'
• The results and running logs will be saved in 'results/'

Others

• If you find our code useful for your research, please give us a star.
• We don't adopt the default settings from the literature. Instead, all stegeanalysis networks are optimized using Adam solver with a weight decay of 1e-5 and an initial learning rate of 2e-4.