This is a project in which side-channel attacks are researched and developed.

The software includes a number of features.

Side-channel attacks included in current version:

• Template attack
• Pooled template attack
• Stochastic attack
• Mutual information analysis

Analysis tools included in current version:

• Normalized inter class variance analysis
• Pearson correlation analysis

Other features include:

• Progressbar
• Logging of sub result

This project needs to following to run:

• Python 3.7.x
• A CUDA enabled grapics card with Compute Capability >= 3.5
  • You can check the CC of your card here: https://developer.nvidia.com/cuda-gpus
• The CUDA SDK (CUDA 10.x)

In a terminal, run

python setup.py install

in the root directory (cs-11/ most likely) to install most of the required dependencies.

Python can't find our modules by itself, so we need to tell Python where they are. The command below will append the current directory (which should be the project root directory) to the PYTHONPATH environment variable.

On Linux or if using a Linux shell on Windows:

export PYTHONPATH=$(pwd)

On Windows' cmd:

set PYTHONPATH=%cd%

Then you can run the program:

python sca

To verify the correctness of your installation, the test suite can be executed with the following command:

python setup.py test

The side-channel attack (sca) program allows the following side-channel attacks to be run:

• Mutual Information Analysis Attack (mia)
• Stochastic Attack (sa)
• (Pooled) Template Attack (ta)
• Online Correlation Power Analysis (cpa-online)

These simple attacks can be run using the following commands

python sca mia
python sca sa
python sca ta
python sca cpa-online

Running just the simple mia attack on the default traces will not always yield the same subkey guess. This behaviour is caused by the number of traces that are used to run the attack. Increasing the number of traces will give more consistent results

python sca mia --traces 2000

Running an attack with custom numpy datasets can be done as follows

python sca ta -tf data/custom_traces.npy -kf data/custom_keys.npy -pf data/custom_plain.npy

It is possible to attack the AES implementation on different encryption rounds as well as the operation to attack.

This example shows how to attack AES on the shift-rows operation on round three using stochastic attack.

python sca sa --op-shift-rows --round 3

The program provides multiple feature selection tools to use. The program also allows you to run an attack without any feature selection tool. However, not using a feature selection tool will sometimes increase the computation time drastically.

This example shows how to use the pearson feature selection instead of the default SOST feature selection.

python sca mia --fs-pearson

If you use this code, please consider citing:

@misc{SCAT,
  author = {Wolfgang Bubberman and Sengim Karayalcin and Matthias Meester and Olaf Braakman and Stjepan Picek},
  title  = {{Side-channel Analysis Toolbox}},
  note   = {{\url{https://github.com/AISyLab/side-channel-analysis-toolbox}}},
  year   = {2020}
}