Side Channel Attacks: Cracking AES with Deep Learning


Cryptosystems are collections of cryptographic techniques used to obfuscate sensitive information by encrypting plaintext into ciphertext and decrypting ciphertext into plaintext for the purpose of secure communication. This project seeks to, on a hardware level, crack/break the AES cryptosystem by performing a deep learning side channel attack. We aim to train a ResNet, LSTM, and transformer to perform such an attack, and evaluate their effectiveness. Ideally, we should be able to collect our own training data.

We are looking for developers who have taken CSC311 or have similar machine learning experience, including familiarity with PyTorch, Numpy/Scipy, as well as developers who have experience with embedded devices and microcontrollers. Developers should also be willing to learn about embedded security. Email or message @whombst on Discord if you’re interested!


The Team