Ensembles for Deep Learning-based Profiled Side-Channel Attacks

Repository code to support paper TCHES2020 (issue 4) paper "Strength in Numbers: Improving Generalization with Ensembles in Machine Learning-based Profiled Side-channel Analysis"

Link to the paper: https://tches.iacr.org/index.php/TCHES/article/view/8686

Authors: Guilherme Perin (Delft University of Technology, The Netherlands), Łukasz Chmielewski (Radboud University Nijmegen and Riscure BV) and Stjepan Picek (Delft University of Technology, The Netherlands)

Datasets

The source code is prepared for three datasets: CHES CTF, ASCAD FIXED KEYS, ASCAD RANDOM KEYS.

CHES CTF dataset

This dataset contains 45,000 profiling traces, with a fixed key, and additional 5,000 attacking traces with a different and fixed key. Each trace contains 2,200 samples that represent the processing of s-box operations around the first AES encryption round (Original CHES CTF webpage https://chesctf.riscure.com/2018/content?show=training provides only 10k traces for each device. The traces are already normalized with z-score normalization).

Download

CHES CTF dataset (ches_ctf.h5 file) can be downloaded from: https://www.dropbox.com/s/lpw1k3so99krmmq/ches_ctf.h5?dl=0

ASCAD FIXED KEY dataset

Information about ASCAD FIXED KEY dataset can be found in the original github page: https://github.com/ANSSI-FR/ASCAD/tree/master/ATMEGA_AES_v1/ATM_AES_v1_fixed_key

ASCAD RANDOM KEY dataset

Information about ASCAD RANDOM KEY dataset can be found in the original github page: https://github.com/ANSSI-FR/ASCAD/tree/master/ATMEGA_AES_v1/ATM_AES_v1_variable_key

Code

This section provides explanations on how to run the ensemble SCA code on the above datasets.

Code execution

The ensemble code should be executed from run_ensemble.py file. This file contain the following structure:

ensemble_aes = EnsembleAES()
ensemble_aes.set_dataset("ches_ctf")  # "ascad_fixed_key", "ascad_random_key" or "ches_ctf"
ensemble_aes.set_leakage_model("HW")
ensemble_aes.set_target_byte(0)
ensemble_aes.set_mini_batch(400)
ensemble_aes.set_epochs(10)
ensemble_aes.run_ensemble(
    number_of_models=50,
    number_of_best_models=10
)

In the example above, the analysis will generate 50 models and create ensembles from the 50 models and from the 10 best models. After run_ensemble method is finished, the user can plot guessing entropy and success rate for ensembles all models, ensembles best models, best validation model and best attack model. As an example, the following code can be used to generate the plot:

import matplotlib.pyplot as plt
plt.subplot(1, 2, 1)
plt.plot(ensemble_aes.get_ge_best_model_validation(), label="GE best validation")
plt.plot(ensemble_aes.get_ge_best_model_attack(), label="GE best attack")
plt.plot(ensemble_aes.get_ge_ensemble(), label="GE Ensemble All Models")
plt.plot(ensemble_aes.get_ge_ensemble_best_models(), label="GE Ensemble Best Models")
plt.xlabel("Traces")
plt.ylabel("Guessing Entropy")
plt.legend()
plt.subplot(1, 2, 2)
plt.plot(ensemble_aes.get_sr_best_model_validation(), label="SR best validation")
plt.plot(ensemble_aes.get_sr_best_model_attack(), label="SR best attack")
plt.plot(ensemble_aes.get_sr_ensemble(), label="SR Ensemble All Models")
plt.plot(ensemble_aes.get_sr_ensemble_best_models(), label="SR Ensemble Best Models")
plt.xlabel("Traces")
plt.ylabel("Success Rate")
plt.legend()
plt.show()

Neural Networks

The provided code generate MLP or CNN with random hyperparameters according to certain user-defined ranges. In commons/ensemble_aes.py, the user can find the methods run_mlp() and run_cnn() for random MLPs and random CNNs, respectively. In commons/neural_networks.py we provide the structure for the neural networks.