steganography-with-masked-lm

This repository contains implementations of the steganography algorithms with masked language models. Please refer to our paper Frustratingly Easy Edit-based Linguistic Steganography with a Masked Language Model for more details.

Requirements

• Python >= 3.6.5

Installation

$ git clone https://github.com/ku-nlp/steganography-with-masked-lm.git
$ cd steganography-with-masked-lm
$ pip install -r requirements.txt

Then, download the NLTK stopwords corpus data (see also https://www.nltk.org/data.html#interactive-installer ).

$ python
>> import nltk
>> nltk.download()
Downloader> d
  Identifier> stopwords

Example usage

$ python main.py "The quick brown fox jumps over the lazy dog." -m 010101
{'stego_text': 'The quick red fox jumps over the poor dog.', 'encoded_message': '010101'}

$ python main.py "The quick red fox jumps over the poor dog." --decode
{'decoded_message': '010101'}

Run python main.py -h for more help.