Trusted-Relay Quantum Key Distribution Network

Session

From

Attacker controls relay

Fiber network

Link distance scale 1.00×

Each real link's loss follows η = 10^(−0.2·L/10) (standard telecom fiber). Stretch the network and long hops go dark — that's why real QKD backbones need relays every ~50–150 km.

Session status

Pick endpoints and run a session.

Path

–

Total distance

–

Session key (Alice's original)

Key recovered at current hop

How trusted-relay QKD works

Point-to-point QKD (BB84, E91, SARG04…) only works over limited distance: photon loss in fiber grows exponentially with length, so past roughly 150–200 km the surviving-photon rate is too low to sift a usable key in reasonable time.

η(L) = 10^(−α·L / 10)     α = 0.2 dB/km (telecom fiber)

Real deployed QKD networks (SECOQC, the Tokyo QKD network, China's Beijing–Shanghai backbone) solve this with trusted relay nodes: every link is short enough to run its own point-to-point QKD session, and a session key hops node-to-node using a one-time-pad relay:

C = K_session ⊕ K_hop        (sender encrypts with the fresh link key)
K_session = C ⊕ K_hop        (receiver, holding the same K_hop, recovers it)

Each relay must hold the session key in the clear for an instant to re-encrypt it for the next hop. That's the well-known weakness of this architecture: a single compromised relay anywhere on the path recovers the entire session key, breaking end-to-end security — unlike a (still largely experimental) quantum-repeater network, which entangles endpoints directly and never exposes the key at an intermediate node.

Link distance scale — stretches every fiber run; watch links cross the loss threshold and turn dead, forcing a reroute or leaving no path at all.
Attacker controls relay — marks one relay as compromised; if the chosen route passes through it, the session is flagged broken even though the key still mathematically arrives intact.

Endpoint (Alice / Bob) Trusted relay Compromised relay Live fiber link Link too lossy to use