Quantum Science for Cybersecurity in Telecommunication Networks: Modelling, Analysis, Simulation and Optimization of QKD Protocols.

Real-world implementations of Quantum Key Distribution (QKD) protocols -- such as the standard BB84 protocol (Bennett and Brassard, 1984) -- face several issues, including eavesdropping attacks, channel losses as well as attenuation phenomena induced by distance. Moreover, random basis selection in original BB84 algorithm affects its pipeline, inherently setting the sifting match at 50%. Optimization of protocol designs and adaptive resources allocation (whether computational or monetary) serve as crucial motivations for approaching such issues through modelling and designing adequate computational solutions. Consequently, this thesis work proposes a Python simulation framework, featuring two models. The first serves as a benchmark, depicting a representation of the standard BB84, its correlated issues and respective solutions. The most structured model, however, is an advanced variant to the original, in which Tree Parity Machines (TPMs) get interfaced with PPO agents behaving in accord with rules extracted from a custom Gymnasium Environment, to ensure users synchronization. Proximal Policy Optimization (PPO) agents are configured to explore various scenarios (induced by variations in channel parameters). The whole model was tested and compared to the baseline benchmark through many experiments, varying the channel length. Indeed, both the models engines drew values from a fixed, finite list, whose equally spaced distances ranged from 0 to 150 km. The scenario focuses on the following: two users, namely Alice and Bob, before the transmission takes place agree onto using a certain percentage of encryption bases per exchange or simulation. This modification proved to be advantageous, improving the sifting rate match, but environmental conditions such as noise challenged the channel transmission, lowering its efficiency upon increasing the distance the information had to cover. Nevertheless, the developed model provides an accessible, versatile tool, executable on any classical device with standard desktop hardware: its implementation guidelines and script files are open-source, available for consultation on a dedicated online repository.