Hound: Locating Cryptographic Primitives in Desynchronized Side-Channel Traces Using Deep Learning

Hound is a tool for locating cryptographic primitives in desynchronized side-channel traces using deep learning. This repository contains the code and resources needed to run experiments and demos.

To run an experiment, open the Jupyter Notebook Hound.ipynb.

Repository Organization

The repository is organized as follows:

• /CNN: Contains modules and configuration files for the Convolutional Neural Network used for classifying the start of cryptographic primitives.
• /inference_pipeline: Contains functions for classifying, segmenting, and aligning cryptographic primitives in a side-channel trace.
• Hound.ipynb: A Jupyter Notebook for running a demo.

.
├── CNN
│   ├── configs
│   │   ├── common
│   │   │   └── neptune_configs.yaml
│   │   └── exp
│   │       ├── data.yaml
│   │       ├── experiment.yaml
│   │       └── module.yaml
│   ├── datasets
│   │   └── cp_class_dataset.py
│   ├── models
│   │   ├── custom_layers.py
│   │   ├── resnet.py
│   │   └── resnet_time_series_classifier.py
│   ├── modules
│   │   ├── cp_class_datamodule.py
│   │   └── cp_class_module.py
│   ├── train.py
│   └── utils
│       ├── data.py
│       ├── logging.py
│       ├── module.py
│       ├── trainer.py
│       └── utils.py
├── inference_pipeline
│   ├── alignment.py
│   ├── debug.py
│   ├── screening.py
│   └── sliding_window_classification.py
└── Hound.ipynb

Dataset

The dataset for the AES cryptosystem is available here on Zenodo.

The dataset is organized as follows:

• /training: Contains three subsets: train, valid, and test. Each subset consists of two .npy files:
  • _set: Contains the preprocessed side-channel traces.
  • _labels: Contains the target labels for training the CNN, labeling each data as CP start, CP spare, or noise.
• /inference: Contains files for two demos: consecutive AES executions and AES executions interleaved with noisy applications. Each demo consists of two .npy files:
  • aes_: Contains the side-channel traces to input into Hound.
  • gt_: Contains the ground truth for checking the correctness of Hound segmentation.

Note

This work is part of [1], available online.

This repository is protected by copyright and licensed under the Apache-2.0 license.

© 2024 hardware-fab

[1] Galli, D., Chiari, G., and Zoni, D. (2024). Hound: Locating Cryptographic Primitives in Desynchronized Side-Channel Traces Using Deep-Learning. Accepted for presentation at the 42nd IEEE International Conference on Computer Design (ICCD). arXiv preprint arXiv:2408.06296.