This repository contains the proof-of-concept implementations and post-processing scripts for the paper SpectrEM: Exploiting Electromagnetic Emanations During Transient Execution.
The directory poc/ contains the proof-of-concept (POC) implementations used to evaluate the performance of the attacks as performed in Section 6. The construction of these POCs is discussed in detail in Section 5.
The scripts used to evaluate the recorded traces are located in scripts/. These Jupyter notebooks walk through the steps we used to train the MLP network, and evaluate a trace using both evaluation methods. This directory also contains scripts that will programmatically go through all evaluation traces to reproduce the results in our paper.
The directory traces/ contains more information about using either the pre-recorded traces or collecting your own traces. The scripts included in this repository will assume the traces are located within this directory.
We provide a repository of pre-recorded traces to enable the reproduction of our results without requiring a similar EM side-channel setup as was used in the paper. These pre-recorded traces cover all experiments in Sections 6 through 8 as well as Appendix C.
Instructions detailing how to download and use these traces can be found here.
To run the Python scripts in this repository, we require the following system resources:
- Memory
To run all scripts, we recommend at least 16 GB of RAM. The exact memory usage of each script can be found here. - Storage
To download and extract all traces, at least 520 GB of free space is required. Each dataset can, however, be downloaded seperately. The download size and uncompressed size for each dataset can be found here.
To get started, first create a new Python environment (python>=3.11) and install all dependencies. For instance, when using Anaconda:
$ conda create -n spectrem "python>=3.11" ipython
$ conda activate spectrem
$ conda config --env --add channels conda-forge
$ conda config --env --set channel_priority strict
$ conda install --file ./requirements.txt
$ pip install tensorflowTo start the Jupyter notebook server:
$ conda activate spectrem
$ jupyter notebookTo get started with the evaluation, visit this notebook which explains how to access the traces and how to evaluate them.
More information on how to reproduce our results can be found here.
@inproceedings{demeulemeester2023spectrem,
author = {De Meulemeester, Jesse and Purnal, Antoon and Wouters, Lennert and Beckers, Arthur and Verbauwhede, Ingrid},
title = {{SpectrEM}: Exploiting Electromagnetic Emanations During Transient Execution},
booktitle = {USENIX Security Symposium},
year = {2023},
}