conseal

Python package, containing implementations of modern image steganographic algorithms.

⚠️ The package only simulates the embedding, which is useful for steganalysis research. We do not provide any end-to-end steganography method.

Installation

Simply install the package with pip3

pip3 install conseal

or using the cloned repository

git clone https://github.com/uibk-uncover/conseal/
cd conseal
pip3 install .

Contents

Steganography method	Domain	Reference
F5	JPEG	Reference
nsF5: no-shrinkage F5	JPEG	Reference
EBS: entropy block steganography	JPEG	Reference
UERD: uniform embedding revisited distortion	JPEG	Reference
J-UNIWARD: JPEG-domain universal wavelet relative distortion	JPEG	Reference
LSB: least significant bit	Spatial / JPEG
HILL: high-low-low	Spatial	Reference
HUGO: highly undetectable stego	Spatial	Reference
MiPOD: minimizing the power of optimal detector	Spatial	Reference
S-UNIWARD: spatial-domain universal wavelet relative distortion	spatial	Reference
WOW: wavelet obtained weights	spatial	Reference

Usage

Import the library in Python 3

import conseal as cl

This package currently contains the three JPEG steganography methods J-UNIWARD, UERD, and nsF5. The following examples show how to embed a JPEG cover image cover.jpeg with an embedding rate of 0.4 bits per non-zero AC coefficient (bpnzAC):

• F5

# load cover
jpeg = jpeglib.read_dct("cover.jpeg")

# embed F5 0.4 bpnzAC
jpeg.Y = cl.F5.simulate_single_channel(
    y0=jpeg.Y,
    alpha=0.4,
    seed=12345)

# save result as stego image
jpeg.write_dct("stego.jpeg")

• nsF5

# load cover
jpeg = jpeglib.read_dct("cover.jpeg")

# embed nsF5 0.4 bpnzAC
jpeg.Y = cl.nsF5.simulate_single_channel(
    y0=jpeg.Y,
    alpha=0.4,
    seed=12345)

# save result as stego image
jpeg.write_dct("stego.jpeg")

• EBS

# load cover
jpeg = jpeglib.read_dct("cover.jpeg")

# embed EBS 0.4 bpnzAC
jpeg.Y = cl.ebs.simulate_single_channel(
    y0=jpeg.Y,
    qt=jpeg.qt[0],
    alpha=0.4,
    seed=12345)

# save result as stego image
jpeg.write_dct("stego.jpeg")

• UERD

# load cover
jpeg = jpeglib.read_dct("cover.jpeg")

# embed UERD 0.4
jpeg.Y = cl.uerd.simulate_single_channel(
    y0=jpeg.Y,
    qt=jpeg.qt[0],
    embedding_rate=0.4,
    seed=12345)

# save result as stego image
jpeg.write_dct("stego.jpeg")

• J-UNIWARD

# load cover
im0 = jpeglib.read_spatial("cover.jpeg", jpeglib.JCS_GRAYSCALE)
jpeg = jpeglib.read_dct("cover.jpeg")

# embed J-UNIWARD 0.4
jpeg.Y = cl.juniward.simulate_single_channel(
    x0=im0.spatial[..., 0],
    y0=jpeg.Y,
    qt=jpeg.qt[0],
    alpha=0.4,
    seed=12345)

# save result as stego image
jpeg.write_dct("stego.jpeg")

• HUGO

# load cover
x0 = np.array(Image.open("cover.png"))

# embed HUGO 0.4 bpnzAC
x1 = cl.hugo.simulate_single_channel(
    x0=x0,
    alpha=0.4,
    seed=12345)

# save result as stego image
Image.fromarray(x1).save("stego.png")

• WOW

# load cover
x0 = np.array(Image.open("cover.png"))

# embed WOW 0.4 bpnzAC
x1 = cl.wow.simulate_single_channel(
    x0=x0,
    alpha=0.4,
    seed=12345)

# save result as stego image
Image.fromarray(x1).save("stego.png")

• S-UNIWARD

# load cover
x0 = np.array(Image.open("cover.png"))

# embed S-UNIWARD 0.4 bpnzAC
x1 = cl.suniward.simulate_single_channel(
    x0=x0,
    alpha=0.4,
    seed=12345)

# save result as stego image
Image.fromarray(x1).save("stego.png")

• MiPOD

# load cover
x0 = np.array(Image.open("cover.png"))

# embed MiPOD 0.4 bpnzAC
x1 = cl.mipod.simulate_single_channel(
    x0=x0,
    alpha=0.4,
    seed=12345)

# save result as stego image
Image.fromarray(x1).save("stego.png")

• HILL

# load cover
x0 = np.array(Image.open("cover.png"))

# embed HUGO 0.4 bpnzAC
x1 = cl.hill.simulate_single_channel(
    x0=x0,
    alpha=0.4,
    seed=12345)

# save result as stego image
Image.fromarray(x1).save("stego.png")

Acknowledgements and Disclaimer

Developed by Martin Benes and Benedikt Lorch, University of Innsbruck, 2024.

The J-UNIWARD and nsF5 implementations in this package are based on the original Matlab code provided by the Digital Data Embedding Lab at Binghamton University. We also thank Patrick Bas and Rémi Cogranne for sharing their implementations of UERD and EBS with us.

We have made our best effort to ensure that our implementations produce identical results as the original Matlab implementations. However, it is the user's responsibility to verify this.