Quantum Key Distribution Secured Federated Learning for Channel Estimation and Radar Spectrum Sensing in 6G Networks

Ferhat Ozgur Catak; Murat Kuzlu; Jungwon Seo; Umit Cali

cs.CRMar 5, 2026

Quantum Key Distribution Secured Federated Learning for Channel Estimation and Radar Spectrum Sensing in 6G Networks

Authors: Ferhat Ozgur Catak, Murat Kuzlu, Jungwon Seo, Umit Cali

Organizations: University of Stavanger (UiS), Stavanger, Norway · Old Dominion University, Norfolk, VA, USA · University of York, York, UK

Abstract

This paper presents a federated learning framework secured by quantum key distribution (QKD) for wireless channel estimation and radar spectrum sensing in the next generation networks (NextG or Beyond 6G). A BB84-style protocol abstraction and pairwise additive masking are utilized to train clients' local models (CNN for channel estimation, U-Net for radar segmentation) and upload only masked model updates. The server aggregates without observing plain parameters; an eavesdropper without QKD keys cannot recover individual updates. Experiments show that secure FL achieves NMSE of 0.216 for channel estimation and 92.1% accuracy with 0.72 mIoU for radar sensing. When an eavesdropper is present, QBER rises to $\sim$ 25% and all rounds abort as intended; reconstruction error remains below $10^{-5}$ , confirming correct aggregation.

Figures & tables

Fig. 3: QKD-secured federated learning system architecture. Clients train locally, mask updates with pairwise keys, and upload; server aggregates without seeing plain parameters.

Fig. 4: BB84 QKD protocol flow: state preparation, measurement, sifting, QBER check, and privacy amplification.

Fig. 5: Pairwise additive masking: each client adds masks for

j>i

and subtracts for

j<i

; masks cancel at aggregation.

Fig. 8: One federated learning round with QKD-secured aggregation and QBER-based abort.

Parameter	Value
Source	MATLAB 5G Toolbox (synthesis)
Image size	$612\times 14$
Training samples	1000
Validation samples	500

TABLE I: Channel Estimation Dataset

Parameter	Value
Center frequency ( $f_{c}$ )	2.8 GHz
Sampling frequency ( $f_{s}$ )	61.44 MHz
Pulse repetition frequency	Variable
Pulse width	1 $\mu$ s
Transmit power	25 kW
Antenna gain	32.8 dB

TABLE II: Radar Signal Parameters

Property	Value
Transmitter positions	Random within 2 km $\times$ 2 km
Number of scatterers	30
Direct path	Disabled
Atmospheric conditions	Simulated

TABLE III: Scenario Setup for Channel Modeling (Radar Dataset)

Parameter	Value
Spectrogram size	$256\times 256\times 3$
Classes	Noise, LTE, NR, Radar
Total samples	200
Train/validation split	80%/20%

TABLE IV: Radar Spectrum Sensing Dataset

$(K,\text{mode})$	Final NMSE	Downlink [bytes]	Uplink [bytes]
$(3,\text{plain})$	$5.03\times 10^{-2}$	$9.4\times 10^{4}$	$2.83\times 10^{5}$
$(3,\text{Classical-SA})$	$6.25\times 10^{-2}$	$9.4\times 10^{4}$	$2.83\times 10^{5}$
$(3,\text{QKD-SA})$	$5.10\times 10^{-2}$	$9.4\times 10^{4}$	$2.83\times 10^{5}$
$(10,\text{plain})$	$6.56\times 10^{-2}$	$9.4\times 10^{4}$	$9.42\times 10^{5}$
$(10,\text{QKD-SA})$	$7.13\times 10^{-2}$	$9.4\times 10^{4}$	$9.42\times 10^{5}$
$(20,\text{plain})$	$8.60\times 10^{-2}$	$9.4\times 10^{4}$	$1.88\times 10^{6}$

TABLE V: Channel Estimation—Experiment A: Utility and Communication vs. Number of Clients

$(K,\text{mode})$	Pixel Acc.	mIoU	Downlink [bytes]	Uplink [bytes]
$(3,\text{plain})$	$88.1\%$	$0.65$	$1.24\times 10^{8}$	$3.73\times 10^{8}$
$(3,\text{QKD-SA})$	$89.0\%$	$0.68$	$1.24\times 10^{8}$	$3.73\times 10^{8}$
$(10,\text{plain})$	$66.9\%$	$0.41$	$1.24\times 10^{8}$	$1.24\times 10^{9}$
$(10,\text{QKD-SA})$	$82.9\%$	$0.56$	$1.24\times 10^{8}$	$1.24\times 10^{9}$
$(20,\text{plain})$	$77.5\%$	$0.47$	$1.24\times 10^{8}$	$2.48\times 10^{9}$
$(20,\text{QKD-SA})$	$73.7\%$	$0.44$	$1.24\times 10^{8}$	$2.48\times 10^{9}$

TABLE VI: Radar Sensing—Experiment A: Utility and Communication vs. Number of Clients

Mode	Rounds	Secure	Aborted
Baseline	5	5	0
Secure	5	5	0
Eve (all rounds)	5	0	5

TABLE VII: Channel Estimation—Experiment B: Threat Model

Outcome	Mean QBER	Mean Acc.	Mean mIoU	Notes
SECURE	—	—	—	Aggregation
				proceeds
ABORTED	24.6%	6.96%	0.0174	All rounds
				aborted

TABLE VIII: Radar Sensing—Experiment B: Threat Model Outcomes

Round	Status	QBER	Acc. Retained
1–10	ABORTED	23.0–26.4%	6.96%

TABLE IX: Radar Sensing—Key Abort Effectiveness (Eve Present)

Round	NMSE	Cosine $r$	Pearson $r$
0	0.2261	0.945	0.943
2	0.2170	0.884	0.877
4	0.2156	0.840	0.828

TABLE X: Channel Estimation—Leakage Proxies (Secure Mode, Sample Rounds)

Federated Learning

Federated learning trains shared models across data holders without requiring them to pool their raw data.

See all 943 papers

Communication-Efficient Quantum Federated Learning over Large-Scale Wireless Networks
Mar 1, 2026Shaba Shaon, Christopher G. Brinton, Dinh C. NguyenQuantum Federated Learning Wireless Resource Allocation
ECE Department, The University of Alabama in Huntsville, Huntsville, AL 35899, USA · Elmore Family School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN, 47907, USA
Privacy-Enhanced Zero-Order Federated Learning via xMK-CKKS over Wireless Channels
May 28, 2026Anthony Ayli, Khalil Harris, Jihad Fahs +1Federated Learning Aggregation Homomorphic Encryption
Learnt Attacks on Quantum Key Distribution under Channel Noise and Device Drift
Oct 1, 2026Marcel Mordarski, Benjamin Gras, Abdelrahman Shehata +2
Department of Computing, Imperial College London, London SW7 2AZ, United Kingdom · Department of Electrical and Electronic Engineering, Imperial College London, London SW7 2AZ, United Kingdom

Mar 1, 2026J/K move · Enter open · S save

Communication-Efficient Quantum Federated Learning over Large-Scale Wireless Networks

Shaba Shaon, Christopher G. Brinton, Dinh C. Nguyen

ECE Department, The University of Alabama in Huntsville, Huntsville, AL 35899, USA · Elmore Family School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN, 47907, USA

Quantum Federated Learning Wireless Resource Allocation Wireless Communications Quantum Machine Learning

Open paper AskPDF arXiv

Quantum federated learning (QFL) combines the robust data processing of quantum computing with the privacy-preserving features of federated learning (FL). However, in large-scale wireless networks, optimizing sum-rate is crucial for unlocking the true potential of QFL, facilitating effective model sharing and aggregation as devices compete for limited bandwidth amid dynamic channel conditions and fluctuating power resources. This paper studies a novel sum-rate maximization problem within a muti-channel QFL framework, specifically designed for non-orthogonal multiple access (NOMA)-based large-scale wireless networks. We develop a sum-rate maximization problem by jointly considering quantum device's channel selection and transmit power. Our formulated problem is a non-convex, mixed-integer nonlinear programming (MINLP) challenge that remains non-deterministic polynomial time (NP)-hard even with specified channel selection parameters. The complexity of the problem motivates us to create an effective iterative optimization approach that utilizes the sophisticated quantum approximate optimization algorithm (QAOA) to derive high-quality approximate solutions. Additionally, our study presents the first theoretical exploration of QFL convergence properties under full device participation, rigorously analyzing real-world scenarios with nonconvex loss functions, diverse data distributions, and the effects of quantum shot noise. Extensive simulation results indicate that our multi-channel NOMA-based QFL framework enhances model training and convergence behavior, surpassing conventional algorithms in terms of accuracy and loss. Moreover, our quantum-centric joint optimization approach achieves more than a 100% increase in sum-rate while ensuring rapid convergence, significantly outperforming the state-of-the-arts.

Abstract

Federated Learning

Explore similar work

Communication-Efficient Quantum Federated Learning over Large-Scale Wireless Networks

Privacy-Enhanced Zero-Order Federated Learning via xMK-CKKS over Wireless Channels

Learnt Attacks on Quantum Key Distribution under Channel Noise and Device Drift