Deep learning based image steganalysis in image spatial domain This is keras based implementation using tensorflow backend. This program is written to work on both cpu as well as gpu with slight modification. It is tested on Nvidia GTX 1080 titan gpu.
Dependencies:
- keras
- Tensorflow
- Numpy
- scipy
- Matplotlib
- PIL
- Sklearn
Dataset requirements:
BOSSbase Dataset can be obtained from http://dde.binghamton.edu/download/ImageDB/BOSSbase_1.01.zip
For data hiding any tool available can be used as per avialability
For testing, alethia can be used to generate stego images in batches Alethia can be downloaded from the respective author's github link provided below: https://github.com/daniellerch/aletheia.git Read the documentation of aletheia to find out the respective dependencies to run that
After obtaining stego and cover images you are ready to use the model Make sure to save the fnet.py and other files in same directory where database of stego and cover images are kept or else make the corresponding changes to the paths in the fnet.py
To run the graph using tensorboard:
save the events file in Graph dir and run the following command:
tensorboard --logdir Graph