Overview

This repository contains the code for the paper "SpectroLoc: Cryptographic Operation Localization via Spectrogram Projection and Time-Series Analysis".

Dataset

The dataset required to run the code can be downloaded from: https://drive.google.com/file/d/17iAaZP0QoogScAILAO1w7NdIjDXl4Ocl/

Extract the downloaded file and place the dataset/ directory in the project root before running the experiments.

Requirements

Required Software

Python 3.9 is recommended. The artifact was tested with Python 3.9.23.

Python Packages

Required packages:

• numpy >= 1.21.0
• scipy >= 1.7.0
• matplotlib >= 3.4.0
• claspy >= 0.1.1
• jupyter >= 1.0.0
• pandas >= 1.3.0

Install the dependencies with:

pip install -r requirements.txt

The core implementation is packaged through pyproject.toml. For local development, reuse from another notebook, or running the notebooks in this artifact, install the repository in editable mode after installing the dependencies:

pip install -e .

The notebooks import the reusable code as spectroloc.*, so this editable install step is required unless src/ is otherwise added to PYTHONPATH.

Usage

The repository includes four primary notebooks corresponding to the main experimental results of the paper:

• 1.grid_search.ipynb: runs the analyst-guided seeding mode on software and hardware AES traces under different projection settings. It corresponds to the projection robustness and parameter grid-search study in Section 4.2.

• 2.auto_search.ipynb: runs the automatic repetitive-CO localization pipeline on AES traces, including the main localization experiments, additive-noise robustness evaluation, and target-length sensitivity analysis. It corresponds to Section 4.3.

• 3.segment.ipynb: performs unsupervised change-point segmentation on ECDSA traces, evaluates the detected boundaries against trigger-derived ground truth, and visualizes the segmentation results. It corresponds to Section 4.4.1.

• 4.disint.ipynb: performs motif-based localization on mixed AES/SHA traces, including interleaved-CO localization, projection-window sensitivity, and additive-noise robustness evaluation. It corresponds to Section 4.4.2.

The notebooks keep dataset paths and paper-specific experiment setup. The spectroloc package contains the reusable core methods:

• spectroloc.projection: time-domain and STFT projection helpers.
• spectroloc.self_temp_analyst: analyst-guided projection, template extraction, and single-combination evaluation helpers.
• spectroloc.self_temp_auto: automatic repetitive-operation localization.
• spectroloc.cpd: change point segmentation, evaluation, and plotting helpers.
• spectroloc.motif: motif localization, evaluation, and plotting helpers for mixed traces.
• spectroloc.config: named dataset configuration classes used by the notebooks.

Directory Structure

spectro-loc/
|-- 1.grid_search.ipynb
|-- 2.auto_search.ipynb
|-- 3.segment.ipynb
|-- 4.disint.ipynb
|-- pyproject.toml
|-- requirements.txt
|-- dataset/
|   |-- chameleon_lowpass/
|   |-- em/
|   |-- semi-loc/
|   `-- trace-copilot/
`-- src/
    `-- spectroloc/
  |-- __init__.py
  |-- config.py
  |-- cpd.py
  |-- motif.py
  |-- projection.py
  |-- self_temp_analyst.py
  `-- self_temp_auto.py

The dataset/ directory is not included in the GitHub repository and should be added manually after downloading and extracting the dataset.

License

This project is released under the MIT License.