Explore deep learning techniques for automating power side channel analysis in this 48-minute Black Hat conference talk. Delve into the intersection of cryptanalytic science and signal processing, learning how to apply convolutional neural networks (CNNs) to overcome challenges in side channel attacks. Discover practical demonstrations on AES-128 attacks, points of interest selection, and template analysis. Gain insights into creating effective training datasets, defining hyper-parameters, and improving model generalization. Witness real-world applications, including bypassing misalignment with CNNs, breaking protected ECC, and tackling first-order masked AES implementations. Walk away with key takeaways and resources for further exploration in this cutting-edge field of cybersecurity.
Lowering the Bar - Deep Learning for Side Channel Analysis
Overview
Syllabus
Intro
Power analysis
Example (huge) leakage
Signal processing (demo)
AES-128 first round attack
Points of interest selection
Concept of Template Analysis
The actual process
Deep Learning
Convolutional Neural Networks (CNN)
Creating training/test/validation data sets
Classification
Step 1: Define initial hyper-parameters (demo)
Make sure it's capable of learning
Make it generalize
Key Recovery
Piñata AES-128 with misalignment (demo)
Bypassing Misalignment with CNNS
Breaking protected ECC on Piñata
Breaking AES with First-Order Masking (demo)
1st cool thing
2nd cool thing
I want to learn more!
Key takeaways
Taught by
Black Hat
