Steganodf

A steganography tool for hiding a message in a dataset, such as csv or parquet files.

This tool hides a payload by permuting the rows of the dataset. The is tolerant to modification thanks to a Reed-Solomon code and a Luby-s LT fontain code.

Demo

You can experiment with the Python API using this Google Colab notebook.

Installation

pip install steganodf

Usage

From command line

# Encoding 
steganodf encode -m hello host.csv stegano.csv
steganodf encode -m hello host.parquet stegano.parquet 
steganodf encode -m hello -p password host.parquet stegano.parquet 

# Decoding 
steganodf decode stegano.csv
steganodf decode stegano.csv -p password

From Python

import steganodf 
import polars as pl
 
df = pl.read_csv("https://gist.githubusercontent.com/netj/8836201/raw/6f9306ad21398ea43cba4f7d537619d0e07d5ae3/iris.csv")

new_df = steganodf.encode(df, "made by steganodf", password="secret")

# Extract your message 
message = steganodf.decode(df, password="secret")

Citation

Sacha Schutz, Meganne Souprayen. Watermark tabular datasets with rows permutations and fountain code. TechRxiv. April 28, 2025. DOI: 10.36227/techrxiv.174585796.61215338/v1 Watermark tabular datasets with rows permutations and fountain code computing and processing