Cryptographie homomorphe et transcodage d’image/video dans le domaine chiffré

Abstract : The emergence of new technologies like cloud computing brings new opportunities in information processing. For example it is easy today to send our personal pictures or videos to a remote server (Google Drive, OneDrive …). We can also share this content among the same servers or via social networks and streaming services. However, this personal data is often also available to the service provider, mainly for practical reasons e.g. to configure a video to have the right format regardless of the displayer (smartphone or computer), to share our data with other people, etc. This raises issues of privacy and trust into the service provider. Classical cryptography brings some answers to this kind of issues, yet leaving the problem of handling the encrypted data: e.g., it becomes impossible to reconfigure a video because the provider can no longer “see” it. An alternative solution is “homomorphic encryption”. It is a powerful tool of advanced cryptography which provides the same security as classical cryptography algorithms, but it still allows us to manipulate ciphertexts such their underlying plaintexts are modified. Consequently, it offers a new perspective to service providers since they can continue to process their clients’information without knowing what it contains. This allows them to provide privacy-preserving services and comply with the new General Data Protection Regulation (GDPR). Although it is considered that homomorphic encryption does not have enough maturity due to its large algorithmic complexity, in this thesis, we are trying to show its potential by using it in the context of image and video processing over the encrypted data. In this context, we look at the different algorithms in an image/video encoder (JPEG/H264 and HEVC) and transform them to circuits which can be manipulated by homomorphic encryption schemes. Our main contribution is to propose the first pipeline for an image compression of type JPEG (homomorphic-JPEG) running on end-to-end encrypted pixels. To optimize the management of the encrypted data, we also propose new tools applicable to existing homomorphic encryption schemes over the ring version of lattices. Our approach allows us to maximize the number of slots in some ciphertext and we introduce new functions allowing to handle these slots independently in the encrypted domain. This thesis work also lead to two publications to international conferences as well as the submission of an additional article.
Complete list of metadatas

Cited literature [116 references]  Display  Hide  Download
Contributor : Abes Star <>
Submitted on : Friday, June 7, 2019 - 1:02:34 AM
Last modification on : Monday, February 10, 2020 - 6:12:35 PM


Version validated by the jury (STAR)


  • HAL Id : tel-02150082, version 1



Donald Nokam Kuaté. Cryptographie homomorphe et transcodage d’image/video dans le domaine chiffré. Cryptographie et sécurité [cs.CR]. Université Paris-Saclay, 2018. Français. ⟨NNT : 2018SACLS575⟩. ⟨tel-02150082⟩



Record views


Files downloads