Claudio Pereti

Trusted-node architectures enable practical metropolitan and wide-area Quantum Key Distribution (QKD) deployments by allowing intermediate relays, typically implemented as Key Management Entities (KMEs), to process and forward key material across chains of QKD links. This design enables scalability but rests on a strong assumption: each relay must be trusted not to access, store, or misuse secret key material. In large multi-hop networks, this creates a growing perimeter of trust whose security depends on the physical and operational integrity of every intermediate node. Although link-level QKD provides information-theoretic security, end-to-end confidentiality ultimately relies on this perimeter, making the protection of secret keys contingent on securing all relay nodes along the path. We propose a simple, implementation-friendly mechanism that reduces the trust required from intermediate relays without modifying the optical layer. The core idea is to mask the raw random sequence used for quantum encoding with a short pre-shared endpoint key before any photon is transmitted. Only the endpoints perform the corresponding logical unmasking after sifting and privacy amplification. Intermediate nodes can still execute standard link-level QKD operations (measurement, sifting, error correction, privacy amplification), yet remain cryptographically unable to reconstruct the final end-to-end key. The approach is fully compatible with existing QKD infrastructures and KME-based architectures and requires only software-level modifications. In summary, our method addresses the long-standing tension between link-level quantum secu- rity and end-to-end trust in multi-hop networks: it preserves the operational role of intermediate relays while preventing them from learning the secret they help deliver.