A Hacker's Guide to Reducing Side Channel Attack Surfaces

Deep Learning in Side-Channel Cryptanalysis

Deep learning models have revolutionized power and electromagnetic side-channel attacks, automating key extraction against hardware cryptographic implementations.

Applying AI Explainability to Hardware Leaks

Elie Bursztein demonstrates leveraging modern AI explainability techniques (such as Integrated Gradients and saliency maps) to inspect neural network activations.

Pinpointing Vulnerable Clock Cycles and Instructions

By visualizing which exact power trace timeframes the neural network utilizes to predict cryptographic key bits, engineers can instantly identify leaky code instructions.

Automating Hardware Defense and Hardening

Bursztein outlines automated testing pipelines for cryptographic developers to eliminate side-channel emissions before silicon fabrication.