Vers la distribution quantique de clé indépendante des dispositifs avec des circuits optiques

Device-Independent Quantum Key Distribution (DI-QKD) is a promising paradigm to secure sensitive communications. This protocol has security guarantees independent of the inner workings of the quantum devices. However, its implementation is challenging, requiring highly entangled states. For now, only one experiment has succeeded in generating such a key with this protocol. It is based on trapped ions and is complex to implement. The objective of this thesis is to develop theoretical tools to reduce the constraints for the application of this protocol, and to apply it to optical devices, which is a promising platform with its high repetition rate and commercially available hardware. In order to do this, we establish a security proof based on all the statistics produced by the setup, which is different from previous methods that base their security on a specific Bell inequality. After this, we present a complete model of a promising optical setup with noise models, then we proceed to analyze on it the performance of the developed protocol. As well, with these new theoretical tools, we apply this analysis on a superconducting platform that has already proven to be adapted for violation of CHSH inequalities, being a figure of merit for the application of DI-QKD. Finally, we apply a reinforcement learning algorithm to search optical circuits with homodyne measurements, maximizing the CHSH score.