MilQAuth: An Improved and Quantum‐Resistant Authentication Framework for Quantum Federated Learning in Military Communication Networks

ABSTRACT Background Federated learning (FL) has emerged as a promising paradigm for distributed model training without centralizing raw data, yet it remains vulnerable to gradient leakage, inference attacks, and malicious updates. The integration of quantum technologies into FL, as proposed by Li et al. through two protocols for quantum federated learning (QFL), aims to mitigate these risks. However, a detailed cryptanalysis of their secure inner‐product estimation and incremental learning protocols reveals critical weaknesses, including susceptibility to replay attacks, quantum tomography, entanglement manipulation, and gradient inversion. These vulnerabilities become particularly severe in mission‐critical environments such as military communication networks, where compromised information may directly expose strategic operations and intelligence. Objective To address these limitations, we propose MilQAuth, an improved quantum‐resistant authentication framework for QFL in military communication networks. Methods The MilQAuth introduces a layered architecture that integrates lightweight authentication, privacy‐preserving encryption, and resilience against both classical and quantum adversaries. Formal validation is provided through the Real‐or‐Random (RoR) model and BAN logic, while automated security verification is conducted using Scyther. Results A comparative performance evaluation demonstrates that MilQAuth achieves reduced computational and communication costs while ensuring stronger privacy guarantees than existing QFL protocols. Conclusions The results confirm that the proposed solution, MilQAuth, not only resists known quantum‐era attacks but also satisfies the stringent security requirements of military‐grade communication systems. Beyond defense applications, the framework remains adaptable to other critical infrastructures, including healthcare and industrial IoT, where secure and privacy‐preserving federated learning is essential.