Implementations of Side-Channel Autoencoders

Refer: Improving Non-Profiled Side-Channel Attacks using Autoencoder-based Preprocessing [pdf]

Non-Profiled Deep Learning-based Side-Channel Preprocessing with Autoencoders [pdf]

Contact: [email protected]

This paper introduces a novel approach with deep learning for improving side-channel attacks, especially in a non-profiling scenario. It propose a new principle of training that trains autoencoders using noise-reduced labels. It notably diminishes the noise in measurements by modifying the autoencoder framework to the signal preprocessing.

The sample source code for the proposed methods in this paper is in the SCAE folder. Also, the code for side channel attacks using side-channel autoencoders is in the Appendix folder. The structure of the source code is as follows:

• SCAE

  • SCAE_Denoise.py (in Section 3-2)
  • SCAE_Align.py (in Section 3-3)
  • SCAE_DK.py (in Section 3-4)
  • hyperparameters.py

• Appendix

  • main.py
  • DDLA.py (in Appendix A) [refer]
  • SCAE_CPA.py (in Appendix A)
  • _SCAE_DDLA.py (Test)
  • loadh5.py
  • hyperparameters.py