Cross-Layer Network Security in the Post-Quantum Era:From Integrated Quantum Sources to Datacenter-Scale Cryptography
Year: 2026
Type: other
Abstract: The security of digital communication rests on mathematical problems that are easy to pose but, for classical computers, practically impossible to solve. The prospect of a cryptographically relevant quantum computer (CRQC) undermines this foundation: once such a machine exists, the public-key cryptography that protects nearly all modern infrastructure can be broken. Although the precise arrival of "Q-Day" is uncertain, the threat is already operational through "Harvest Now, Decrypt Later" (HNDL) attacks, in which adversaries store encrypted traffic today to decrypt it once the hardware matures. Mitigating this threat is not a single algorithm-selection problem but a tiered, heterogeneous re-engineering of communication systems. This shift is driven by fundamental trade-offs between security guarantees, deployment complexity, and performance: computationally secure post-quantum cryptography (PQC) scales to protect high-volume datacenter traffic but relies on mathematical assumptions, whereas quantum key distribution (QKD) offers information-theoretic, physics-based security at the cost of specialized optical hardware and dedicated infrastructure. A comprehensive defence must therefore integrate both paradigms. The central premise of this dissertation is that scaling quantum-resilient communication is an integration challenge governed by boundaries among the device, protocol, and infrastructure layers. A cryptographic primitive that is secure in isolation may still be impractical if it cannot meet the latency, throughput, determinism, power, or observability constraints of the network in which it runs; likewise, a quantum primitive becomes operational infrastructure only when its photonic devices, post-processing pipeline, and classical integration points are engineered together. The work was carried out within the Quantum Resistant Communications: Advance Learning in Applied PQC Training (QUARC) doctoral network as an industrial doctorate embedded with NVIDIA's networking organization. The dissertation comprises peer-reviewed publications, manuscripts under review, and patent applications grouped into three parts. Part I treats QKD as a deployable network capability: it presents the design and characterization of integrated photonic entanglement sources for wavelength-division-multiplexed QKD, and the offload of QKD post-processing onto data processing units (DPUs). Part II investigates how PQC algorithms can be executed efficiently on heterogeneous CPU--GPU--DPU platforms and applied to secure end-to-end real-time pipelines. Part III examines datacenter-scale security, addressing the ordering- and timing-related risks of RDMA over high-throughput optical fabrics and cross-layer observability and intrusion detection for confidential artificial intelligence (AI) infrastructure. Taken together, these contributions show that device-level, protocol-level, and infrastructure-level design choices are tightly coupled: decisions made at one layer constrain or enable the others. The synthesis argues that quantum-resilient communication is achievable with today's and near-future technology, but only by actively engineering the integration points where cryptography, photonics, and programmable network hardware meet, rather than treating them as independent disciplines.
Source: TU/e Research Portal
Language: English
Subfield: Atomic and Molecular Physics, and Optics
Field: Physics and Astronomy
Domain: Physical Sciences
Open Access status: green
Awards: G2126599707, G7946330863
OpenAlex ID: W7221383788