Summary Table of Differential Privacy for Wireless Federated Learning (2026)
Published:
| #Num | References | Wide/narrow band | MAC model | Convergence (assumption) | Optimization variable |
|---|---|---|---|---|---|
| 1 | “Wireless federated learning with local differential privacy”, ISIT, 2020 | time-invariant flat-fading | NOMA | smooth, strongly convex, bounded sample-wise gradient | n/a |
| 2 | “Differentially private aircomp federated learning with power adaptation harnessing receiver noise”, Globecom, 2020 | time-invariant flat-fading | NOMA | analysis SNR-privacy trade-off | power-scaling factor |
| 3 | “Privacy for free: Wireless federated learning via uncoded transmission with adaptive power control”, JSAC, 2020 | block flat-fading | OMA and NOMA | smooth, PL | power-scaling factor, artificial noise scale |
| 4 | “Privacy amplification for federated learning via user sampling and wireless aggregation” | block flat-fading | NOMA | smooth, strongly convex, bounded sample-wise gradient | |
| 5 | “Private wireless federated learning with anonymous over-the-air computation”, ICASSP, 2021 | block flat-fading | NOMA | n/a | n/a |
| 6 | “Device scheduling for over-the-air federated learning with differential privacy”, ICC, 2023time-invariant flat-fading | time-invariant flat-fading | NOMA | smooth, PL, bounded sample-wise gradient | #scheduled devices set, alignment coefficient |
| 7 | “Device Scheduling for Secure Aggregation in Wireless Federated Learning”, IOTJ, 2024 | block flat-fading | NOMA | smooth, strongly convex, bounded sample-wise gradient | device scheduling strategty (joint training or send noise) |
| 8 | “Towards differentially private over-the-air federated learning via device sampling”, Globecom, 2023 | time-invariant flat-fading | NOMA | not a standard convergence result, convex, smooth | #sampling devices set, transceiver design factor |
| 9 | “Communication-learning co-design for differentially private over-the-air federated learning with device sampling”, TWC, 2024 | time-invariant flat-fading | NOMA | not a standard convergence result, convex, smooth (also non-convex and multiple local iterations cases) | #sampling devices set, transceiver design factor, #iterations |
| 10 | “On the differential privacy in federated learning based on over-the-air computation”, TWC, 2023 | block flat-fading | NOMA | smooth | n/a, need parameter estimation |
| 11 | “On the privacy leakage of over-the-air federated learning over MIMO fading channels”, Globecom, 2023 | time-invariant flat-fading | NOMA, MIMO | n/a | private information extraction optimization through receiver beamformer |
| 12 | “Differentially private over-the-air federated learning over MIMO fading channels”, TWC, 2024 | time-invariant flat-fading | NOMA, MIMO | smooth, strongly convex, bounded sample-wise gradient | (1) use $M$ parallel linear estimateors to extract private information; (2) receive combiner, normalizing factor, power fraction factors |