Hacking the privacy amplification of quantum key distribution with machine learning, and countermeasures: An argument for considering classical side channels in quantum key distribution

12/05/2022

Quantum key distribution (QKD) exploits the principles of quantum mechanics to generate and distribute private keys using quantum systems and an authenticated public classical channel. Although it offers information-theoretical security, its physical implementations usually do not due to unintended sources of information leakage called side channels. Side channels in QKD are either similar to those found in classical implementations or related to the quantum processing. Current research and development have often neglected the former. In this thesis, we briefly review the main breakthroughs in QKD security analysis, the main components of QKD implementations and algorithms for classical postprocessing and the use of repeaters for extending transmission distance. We also comprehensively review the main side-channel attacks demonstrated in QKD implementations and their countermeasures. Moreover, we propose a classical-side-channel attack on the privacy-amplification step of a general QKD protocol based on matrix hashing, using machine-learning techniques to analyse its power-consumption leakage. We analyse multiple simulated scenarios and are often able to recover the full private key. Gradient-boosting machine was the best-performing model in virtually every scenario, recovering the full key for high-enough measuring-instrument sampling rates with any hashing-matrix size and noise level tested. In case of a non-perfect model, we devise a strategy, based on analysing confusion matrices, which can make a brute-force search for the key feasible. We also discuss countermeasures based on noise insertion, masking and randomization techniques. This work demonstrates that machine-learning techniques can be used to robustly characterize the leakages in a QKD implementation and generate powerful attacks.

Autores da comunidade :

• 

  J

  João Diogo Ferreira Bravo

  ist425690

Orientadores desta instituição:

• 

  J

  João Carlos Carvalho de Sá Seixas

  ist11944

• 

  P

  Paulo Alexandre Carreira Mateus

  ist13783

physical-sciences - Física

por - Português

Instituto Superior Técnico