arXiv:2610.26120
Multi-path Security in Quantum Networks with Trusted Nodes
Muhammad Kumail Haider, Aeysha Khalique
Abstract
Quantum communication networks promise information-theoretic security but this security collapses if intermediate relay nodes are compromised, a risk that grows sharply when compromised nodes cannot be identified. We address this twofold: first, we show that multi-path routing provides quantifiable protection against unidentified compromised nodes. Second, we provide a comprehensive statistical and graph-theoretic analysis of network and node security under compromise, covering full and partial node compromise, optimal placement of additional edges via an iterative method, and comparative hacking-probability analysis across network topologies. Together, these results identify the topology and multi-path routing strategy that most improves security for a given network. We apply this framework to quantum key distribution across multiple network topologies, demonstrating a scalable route to secure, reliable quantum communication under realistic, unidentified adversarial threats, without relying on quantum memory.
AI Summary
The paper studies how multipath routing can protect trusted-relay QKD networks when compromised intermediate nodes cannot be identified. It assumes secure endpoints and combines independently distributed path keys by XOR, so the final key remains secret whenever at least one constituent path is uncompromised. For independently compromised nodes with security probabilities $p_v$, a path has security probability $P_{\mathrm{path}}=\prod_v p_v$. For internally vertex-disjoint paths, the model gives $P_{\mathrm{net}}=1-\prod_i(1-P_{\mathrm{path},i})$. Routing uses negative-logarithmic node weights and repeated shortest-path searches, deleting previously used intermediate nodes; the stated runtime is $O(T|V|^2)$ for $T$ paths. A separate construction enumerates all $\binom{|V|-2}{T}$ possible sets of compromised nodes and collects paths avoiding each set, providing protection against up to $T$ compromises when every set admits an avoiding path. Numerical comparisons cover seven ten-node topologies: line, star, cycle, grid, tree, wheel, and connected tree. They evaluate security versus node-compromise probability, enumerate fixed-size compromise sets to calculate pairwise hacking probabilities, and search candidate additional edges for improvements in network-average security. Multipath gains appear in topologies supporting redundant routes, whereas line, star, and tree networks show no such gains. The QKD application models the final rate as $R_{\mathrm{final}}=P_{\mathrm{net}}\min_i R_{\mathrm{path},i}$ and reports increased modeled rates with additional paths. The stated computational limitation is the combinatorial growth of the compromise-enumeration construction. The probabilistic analysis assumes independent node compromises; its topology comparisons use ten-node networks.
Central claim may not hold: GreedyPaths() can destroy the possibility of finding two internally disjoint paths even when they exist, contradicting its claimed feasibility and security optimality. The exhaustive compromise-set and simple-path analyses have exponential worst-case scaling, so the ten-node demonstrations do not establish the advertised general scalability. The statistical formula correctly gives the probability that at least one path is secure under its stated assumptions, but the accompanying sentence incorrectly calls it the probability that all paths are secure. The conclusion's endpoint-degree condition for guaranteed security is false because a smaller internal vertex cut can compromise every route. The proposed final key rate multiplies throughput by a probability of secrecy without establishing that the resulting output satisfies a cryptographic secrecy criterion.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 20 pages