Autor: |
Zhou, Yuanyuan |
Přispěvatelé: |
UCL - SST/ICTM/ELEN - Pôle en ingénierie électrique, UCL - Ecole Polytechnique de Louvain, Standaert, François-Xavier, De Vleeschouwer, Christophe, Gierlichs, Benedikt, Koeune, François, Cagli, Eleonora |
Jazyk: |
angličtina |
Rok vydání: |
2023 |
Předmět: |
|
Popis: |
Side-channel analysis (SCA) is one pillar of security evaluations of cryptographic algorithm implementations. Due to different widely used SCA countermeasures and new evaluation requirements of using more side-channel traces and applying deep learning techniques, more efficient side-channel information extraction and exploitation techniques are essential for security evaluations. In this thesis, I investigate such techniques in the context of SCA security evaluations. The first part is committed to traditional multivariate techniques. Concretely, I contribute to i) a worst-case horizontal attack evaluation of elliptic curve cryptosystems (ECC) by extracting information as much as possible, ii) comparing the state-of-the-art Scatter technique against misalignment countermeasures with the linear regression technique on symmetric cryptography from the information extraction perspective. The second part focuses on deep learning-based information extraction. More precisely, I first study the impact of alignments on deep learning SCA evaluations. Second, I simplify the previous worst-case horizontal attack on ECC thanks to the deep learning technique. Third, I propose more efficient profiling attacks via S-box pooling on symmetric cryptography algorithms and deep learning SCA. The third part devotes to information exploitation after the SCA information extraction. Concretely, I put forward the bounds of required SCA key information extraction by extending the partial key exposure attack to CRT-RSA with additive exponent blinding countermeasure. (FSA - Sciences de l'ingénieur) -- UCL, 2023 |
Databáze: |
OpenAIRE |
Externí odkaz: |
|