Deep learning a threat to hardware encryption

Deep learning could be used to crack the encryption codes used to safeguard secure products. Leti’s CESTI* studied the threat in research that recently resulted in a PhD dissertation that earned broad international recognition.
The research focused on attacks via auxiliary channels. These attacks log massive amounts of data on a chip’s electricity consumption and electromagnetic emissions, and then use deep learning algorithms to analyze the data and crack the cryptographic codes.
The PhD dissertation that resulted from the research put forward a type of attack that CESTI can now measure and implement when conducting evaluations. CESTI is now also poised to join France’s National Cybersecurity Agency (ANSSI) in raising developers’ awareness of this new threat.

*Information technology security evaluation center

Contact: eleonora.cagli@cea.fr

Facebook

Twitter