Audio Steganalysis with Deep Learning
@ Author: Yuntao Wang (Charles_wyt)
@ Email: email@example.com
Hope we have a happy communication.
This project is a tensorflow implementation of our recent work, and you can design your own networks through the platform.
- CNN-based Steganalysis of MP3 Steganography in the Entropy
Code Domain [IH&MMSec 2018, Best Paper Award]
[Paper (ACM)] [Paper (pdf)] [Dataset]
- RHFCN: Fully CNN-based Steganalysis of MP3 with Rich High-Pass Filtering [ICASSP 2019]
[Paper (IEEE)] [Paper (pdf)] [Dataset]
You can use the command, pip install -r requirements.txt, to install all necessary packages mentioned above. If you don't want to change or break your original version of tensorflow, you can use virtualenv to create a new python runtime environment.
How to Use
- Install Python3.x or Anaconda, and add the installation directory into the environment variable (recommand python3.5).
- GPU runtime environment configure for the network training (optional).
- Install all dependent packages mentioned above (open setup/requirements.txt, and input "pip install -r requirements" into your terminal window).
- Run the code as the example shows.
- Use tensorboard to visualize the training process such as the accuracy and loss curve of the training. The command is "tensorboard --logdir=/path of log".
- If you want to design your own network based on this project, there is a brief instruction for you.
- All our sourcecode is writen with Pycharm, and the hard wrap is setted as 180. If your setting of hard wrap is less than 180, there will be warnings shwon in the IDE.
|1||audio_samples||some audio samples|
|2||data_processing||dataset build, tools which are used for QMDCT coefficients extraction and so on|
|3||jupyter||a folder for debug with jupyter|
|4||papers||the paper, presentation, dataset and brief introduction of our recent work|
|5||setup||setup and configuration|