tmla 0.1.0

T-MLA: Targeted Multiscale Log-Exponential Attack for Neural Image Compression

T-MLA: A Targeted Multiscale Log-Exponential Attack Framework for Neural Image Compression

Wavelet-aware adversarial perturbations. (a) PGD-based attack with small-magnitude noise, visually clean yet disruptive after compression. (b) Proposed wavelet-aware attack is also imperceptible but more stealthy. (c) Wavelet coefficients of (a) reveal widespread noise in flat regions. (d) Coefficients of (b) closely resemble the clean input, indicating reduced detectability.

Official implementation of "T-MLA: A targeted multiscale log–exponential attack framework for neural image compression" (Information Sciences, Q1).

📄 Abstract

Neural image compression (NIC) has become the state-of-the-art for rate-distortion performance, yet its security vulnerabilities remain significantly less understood than those of classifiers. Existing adversarial attacks on NICs are often naive adaptations of pixel-space methods, overlooking the unique, structured nature of the compression pipeline. In this work, we propose a more advanced class of vulnerabilities by introducing T-MLA, the first targeted multiscale log–exponential attack framework. We introduce adversarial perturbations in the wavelet domain that concentrate on less perceptually salient coefficients, improving the stealth of the attack. Extensive evaluation across multiple state-of-the-art NIC architectures on standard image compression benchmarks reveals a large drop in reconstruction quality while the perturbations remain visually imperceptible. On standard NIC benchmarks, T-MLA achieves targeted degradation of reconstruction quality while improving perturbation imperceptibility (higher PSNR/VIF of the perturbed inputs) compared to PGD-style baselines at comparable attack success, as summarized in our main results. Our findings reveal a critical security flaw at the core of generative and content delivery pipelines.

Installation

Option 1: From GitHub (recommended)

pip install git+https://github.com/nkalmykovsk/tmla.git

Then clone the repo to get scripts/, init.py, and demo.ipynb, and run setup:

git clone https://github.com/nkalmykovsk/tmla.git
cd tmla
python3 init.py   # LIC_TCM + data + weights

Option 2: Clone and install locally

1. Clone the repository

   git clone [e-mail]:nkalmykovsk/tmla.git
   cd tmla
   

2. Environment

   • venv:

     python3 -m venv venv && source ./venv/bin/activate
     pip install -r requirements.txt
     pip install -e .
     

   • Docker:

     docker build -t tmla-lic-tcm:latest .
     docker run -d --gpus all --name tmla-dev \
       -v "$(pwd)":/app -w /app tmla-lic-tcm:latest tail -f /dev/null
     

3. Model and data
   Fetch the LIC-TCM model and datasets (Kodak, DIV2K, CLIC):

   python3 init.py
   

   With Docker: docker exec -it tmla-dev python3 init.py

4. Notebook
   Open demo.ipynb from project root (so tmla is importable).

---

Project structure

Repository = the whole project (this repo, folder tmla/ on disk).
Package = the Python code you import (import tmla); that is the tmla/ directory and its .py files. Having the repo and the package share the same name (tmla) is normal (e.g. numpy, requests).

• tmla/ — Python package: config, attacks (multiscale attack, decomposition, reconstruction, metrics), tcm (loader for LIC_TCM model), utils.
• scripts/ — CLI entry points: run_attack.py, run_batch_parallel.py, compute_entropy.py, collect_metrics.py, build_chart.py.
• init.py — setup script (clones LIC_TCM, downloads data and weights).
• LIC_TCM/ — cloned repo and weights (created by init.py); TCM model code is loaded from here at runtime.

---

Usage

Attacks

• Single image

  python3 scripts/run_attack.py --image path/to/image.png --model model_name
  

• Batch (dataset × model list)

  python3 scripts/run_batch_parallel.py
  

Image complexity (entropy)

Local entropy reflects spatial complexity and is used in the paper for analysis. The script below computes the normalized local Shannon entropy map Ẽ(x,y) ∈ [0, 1] and the global complexity score μ_E.

python3 scripts/compute_entropy.py path/to/image.png --save entropy_map.png --show

---

---

Publishing the package (for maintainers)

The package is ready to publish. Users can already install it from GitHub (see above). To publish on PyPI so that pip install tmla works:

1. Create an account on pypi.org (and test.pypi.org for testing).
2. Install build tools: pip install build twine.
3. From the project root:

   python -m build
   twine upload dist/*
   

   For Test PyPI first: twine upload --repository testpypi dist/*.
4. After publishing, anyone can run: pip install tmla. They will still need to clone the repo and run init.py to get LIC_TCM and datasets (or use the repo as the main way to get scripts and data).

---

Citation

If you use this code or the (paper), please cite:

@article{kalmykov2026tmla,
  title   = {T-MLA: A Targeted Multiscale Log--Exponential Attack Framework for Neural Image Compression},
  author  = {Kalmykov, N. I. and Dibo, R. and Shen, K. and Zhonghan, X. and Phan, A. H. and Liu, Y. and Oseledets, I.},
  journal = {Information Sciences},
  volume  = {702},
  pages   = {123143},
  year    = {2026},
  publisher = {Elsevier},
  doi     = {10.1016/j.ins.2025.123143}
}