Indian Institute of Technology Delhi (IIT Delhi)
National Quantum Mission (NQM) — Quantum Communication T-Hub
Supervisor: Dr. Neel Kanth Kundu, IIT Delhi
Research Intern at IIT Delhi under the National Quantum Mission, designing NARAD-QC: a quantum-network framework that selects, authenticates and executes routes across repeater chains and distils a secret key end to end. Routing is driven by service models calibrated on 640 simulated trajectories and studied through 4,500 Monte Carlo trials. Journal manuscript in preparation with Dr. Neel Kanth Kundu and Dr. Neha Sharma.
How can a quantum communication network dynamically select reliable, authenticated, low-error paths between multiple users and repeaters? That was the question put to a working point-to-point QKD system. NARAD-QC is the attempt to answer it, not with a closed-form argument, but by building a network simulator honest enough to be wrong and then letting its own data decide how routes should be ranked.
Distance is the obvious routing metric and the wrong one. A short route can fail because one fibre generates entanglement too rarely, because a repeater has no free memory, because swaps keep failing, or because stored pairs decay before their neighbours arrive. Adding repeaters shortens each hop but adds swapping and waiting costs. The route that wins depends on all of these at once, and no single physical parameter predicts it.
| Stage | What Was Done |
|---|---|
| Baseline | Rebuilt the point-to-point CrypTon QKD and AES system as the reference. |
| Physics | Added distance-dependent loss, then a single repeater, then a discrete-event engine with arbitrary repeater chains, finite memory, swapping trees and decoherence. |
| Characterisation | Monte Carlo studies of how the engine behaves across distance, repeater density and swap-scheduling policy. |
| Calibration | 640 routed trajectories across 80 network configurations, used to fit service models. |
| Routing | A policy built from the fitted models, with hardware feasibility and QKD quality as hard constraints. |
| Integration | Route discovery, ranking, ML-DSA authentication, entanglement delivery, key distillation and fallback, joined into one pipeline that ends in an encrypted session. |
Swap-scheduling study: 2,400 trials. Eight distances from 100 to 1,000 km, three scheduling policies, 100 trials each, with identical physical randomness for every policy within a trial. All three delivered in 100% of trials through 500 km, and none delivered a pair at 1,000 km. In between there was no universal winner. The static balanced tree is fastest at short range but delivers the lowest-fidelity pairs. Max-parallel delivers the highest-fidelity pairs at nearly every distance. At 700 km, sequential and max-parallel are indistinguishable (56% against 54%) while the static tree collapses to 26%.
Repeater-density sweep: 2,100 trials. At 600 km with 3 to 9 repeaters, success rises from 1% to 94% with a sharp knee between four and six repeaters (5% to 80%), then flattens. Which policy is best changes with density.
Calibration: 640 trajectories. 96.4% delivered a pair. Delivery times were heavy-tailed, with a median of 0.07 s, a mean of 0.64 s and a worst case of 17.5 s. A simple analytic delivery-time proxy tracks the simulated times only moderately (Spearman ρ ≈ 0.72), so the service models are fitted from data rather than trusted from first-principles estimates.
A rule that wins at one distance or density loses at another. That is the case for modelling each path, not trusting a rule of thumb.
A first attempt at implementing a published scheme (RRI) failed and was followed by a second attempt. Hand-set routing weights were replaced by fitted models. A hard reliability admission threshold was tested for the routing policy and dropped because it wasn't stable. The notebook keeps the earlier baselines and experiments alongside the final design.
This is a software simulation with assumed physical parameters, and its key distillation is not a composable finite-key security proof. The routing policy is calibrated within 200–600 km, 2–8 repeaters and 1–16 multiplexing modes, and it has not yet been benchmarked head-to-head against baseline routers. These are the stated boundaries of the work.
The complete system was demonstrated end to end to the supervisor in October 2026. A journal manuscript is in preparation with Dr. Neel Kanth Kundu and Dr. Neha Sharma. The engineering story, including the API, the parallel engine and the React console, is on the NARAD-QC project page.
NARAD-QC : Network Architecture for Routing, Authentication and Delivery in Quantum Communications
Vishesh Goyal, Dr. Neha Sharma, Dr. Neel Kanth Kundu
Journal manuscript in preparation