Rinovative/pd12m_dct_based_synthetic_stegano

PD12M DCT-Based Synthetic Steganography Dataset

This dataset is a synthetic steganographic image dataset based on the PD12M (Public Domain 12M) image collection.
It was designed to simulate detectable modifications similar to those produced by real-world JPEG steganography algorithms, using only public domain data.

For each selected image, three synthetic stego variants were created using targeted DCT (Discrete Cosine Transform) manipulations in the Y channel:

• synthetic_JMiPOD: Random Gaussian noise added to mid-frequency DCT coefficients.
• synthetic_JUNIWARD: Slight scaling of selected high-frequency DCT coefficients.
• synthetic_UERD: Sparse random perturbations applied across the full DCT block.

These modifications simulate detectable steganographic alterations for research purposes but are not real steganographic embeddings.

🔧 Modification Logic

Each image was processed in 8×8 DCT blocks. Example logic:

# JMiPOD-like: Add Gaussian noise to mid-frequency area
jmipod_block[2:6, 2:6] += np.random.normal(0, 0.5, (4, 4))

# JUNIWARD-like: Slightly amplify high frequencies
juniward_block[4:8, 0:4] *= 1.05

# UERD-like: Sparse random perturbation across full block
mask = np.random.rand(8, 8) < 0.05
uerd_block += mask * np.random.normal(0, 1.0, (8, 8))

All operations are applied to the Y channel in the YCrCb color space before JPEG re-encoding.

📂 Dataset Structure

/Cover/                 → Original PD12M images (resized + JPEG recompressed)
/synthetic_JMiPOD/      → DCT-modified images (JMiPOD-inspired)
/synthetic_JUNIWARD/    → DCT-modified images (JUNIWARD-inspired)
/synthetic_UERD/        → DCT-modified images (UERD-inspired)

All images are JPEG-compressed and named using a uniform index (e.g., 00001.jpg).

📜 License and Sources

• Original PD12M Dataset: Public Domain / CC0
  source.plus/pd12m

• This synthetic dataset: CC0

This dataset does not contain real steganographic payloads.
It was created solely for academic and reproducibility purposes.

📚 Citation

If you use this dataset, please reference:

• The original PD12M image dataset (Public Domain)
• This synthetic modification (Rinovative, 2025)

---

📄 Project Reference

Note:
This synthetic dataset was created as a publicly available demonstration subset for the project:
"Deep Learning for Steganalysis – ALASKA2 Dataset"
(GitHub Repository – Rinovative/alaska2-steganalysis)
(BSc Systems Engineering, Computational Engineering Major, OST – Spring 2025, Rino Albertin).

The project primarily uses the ALASKA2 dataset.
In cases where ALASKA2 is not available, this synthetic dataset provides a compatible fallback for running the full pipeline and reproducing the project results.

Important:
This demo subset is fixed and remains consistent with the final project version.