This code belongs to "VAE-Stega: Linguistic Steganography Based on Variational Auto-Encoder".

• python 3
• tensorflow = 1.12
• zhusuan = 0.3.1
• gensim

• Download our corpus file movie or tweet and put it in ./data/
• Modify ./utils/parameters.py to adjust the hyperparameters

• Download our corpus file movie or tweet and put it in ./data/
• Download "BERT-Base, Uncased model" (bert) and put it in ./pre/uncased_L-12_H-768_A-12
• Modify ./utils/parameters_bert.py to adjust the hyperparameters

• VAE_LSTM-LSTM model

python vae_lstm-lstm.py

• VAE_BERT-LSTM model

python vae_bert-lstm.py

This code is based on yiyang92's vae_for_text, google-research's bert and huwenxianglyy's bert-use-demo. Many thanks!