SAMOP 2023 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
QI: Fachverband Quanteninformation
QI 8: Quantum Communication I (joint session Q/QI)
QI 8.2: Talk
Monday, March 6, 2023, 17:15–17:30, F442
Hacking QKD Sender Electronics Using Deep Learning — •Adomas Baliuka1,2, Markus Stöcker1,2, Michael Auer1,2,3, Peter Freiwang1,2, Harald Weinfurter1,2,4, and Lukas Knips1,2,4 — 1Fakultät für Physik, Ludwig-Maximilians-Universität, 80799 München — 2Munich Center for Quantum Science and Technology, 80799 München — 3Universität der Bundeswehr, 85577 Neubiberg — 4Max-Planck-Institut für Quantenoptik, 85748 Garching
Quantum key distribution (QKD) promises provably secure communication. However, the proofs make assumptions which have to be met carefully in practical implementations. Violations of the assumptions open up side channels, which enable an eavesdropper to obtain secret information. For a QKD sender, imperfections in quantum state preparation can lead to quantum side channels by encoding secret information in degrees of freedom (e.g., frequency, spatial mode) not protected by the QKD protocol. On the other hand, information can also leak via classical side channels, such as acoustic vibrations or classical electromagnetic emissions.
We analyze electromagnetic emissions from the electronics of our home-built BB84 QKD sender at a distance of a few centimeters. We are able to extract virtually all information about the secret key using a neural network and even observe traces of electromagnetic radiation at distances of up to a few meters. We discuss countermeasures and evaluate a revised electronics design, showing a significant reduction of emissions and attack performance.