Deep Image Steganography

Image-into-Image Steganography Using Deep Convolutional Network

StegNet: Mega-Image-Steganography-Capacity-with-Deep-Convolutional-Network

---

PCM2018 Paper

PDF Version

Future Internet Paper

PDF Version

HTML Version

How to create ImageNet Dataset used by StegNet

Read the LMDB Creator Doc

How to run the StegNet Model

Step 1. Setup Environmental Variables:

export ILSVRC2012_MDB_PATH="<Your Path to Created 'ILSVRC2012_image_train.mdb' Directory>"

Step 2. Run the code

python ./main.py

The command line arguments can be tweeked:

  -h, --help
  --train_max_epoch TRAIN_MAX_EPOCH
  --batch_size BATCH_SIZE
  --restart                          # Restart from scratch
  --global_mode {train,inference}

Please cite as

PCM 2018

@InProceedings{StegNet-PCM2018,
  AUTHOR = {Wu, Pin and Yang, Yang and Li, Xiaoqiang},
  EDITOR = {Hong, Richang and Cheng, Wen-Huang and Yamasaki, Toshihiko and Wang, Meng and Ngo, Chong-Wah},
  TITLE = {Image-into-Image Steganography Using Deep Convolutional Network},
  BOOKTITLE = {Advances in Multimedia Information Processing -- PCM 2018},
  YEAR = {2018},
  PUBLISHER = {Springer International Publishing},
  ADDRESS = {Cham},
  PAGES = {792--802},
  ISBN = {978-3-030-00767-6}
}

Future Internet

@Article{StegNet-FutureInternet,
  AUTHOR = {Wu, Pin and Yang, Yang and Li, Xiaoqiang},
  TITLE = {StegNet: Mega Image Steganography Capacity with Deep Convolutional Network},
  JOURNAL = {Future Internet},
  VOLUME = {10},
  YEAR = {2018},
  NUMBER = {6},
  ARTICLE NUMBER = {54},
  URL = {http://www.mdpi.com/1999-5903/10/6/54},
  ISSN = {1999-5903},
  DOI = {10.3390/fi10060054}
}