An Adaptive Differentially Private Federated Learning Framework

Fei Xing; Hui Ma; Jin Wang; Ming Yan; Xinjun Pei; Yajun Zhang; Yikun Chen

REVIEW 3 major objections 3 minor 32 references

An Adaptive Differentially Private Federated Learning Framework

T0 review · 3 major / 3 minor · reviewed 2026-08-03 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read FedDimDP claims adaptive clipping, local feature compression, and robust weighting make differentially private federated learning stable and accurate under non-IID data.

desk verdict A clearly written adaptive DP-FL combination, but the privacy guarantee is asserted, not derived—batch-gradient clipping does not bound per-sample sensitivity, so the central claim fails as stated. read the letter →

arxiv 2602.06838 v3 pith:D3RW5MYY submitted 2026-02-06 cs.AI

Jin Wang , Hui Ma , Yajun Zhang , Xinjun Pei , Ming Yan , Fei Xing , Yikun Chen This is my paper · ORCID

classification cs.AI

keywords federatedlearningdifferentialprivacyadaptiveclippingrobustaggregationnon-IIDdatadimensionalityreductiongradientsparsification

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper proposes FedDimDP, a differentially private federated learning framework aimed at the twin problems of non-IID client data and DP noise. It argues that fixed clipping thresholds and naive averaging cause training oscillation and accuracy loss under privacy constraints, and replaces them with three coordinated mechanisms: a local module that projects features into a low-dimensional space and sparsifies gradients, a server-side clipping threshold that adapts to the median of recent update norms, and a robust aggregation rule that down-weights outlier updates while staying near a utility-based nominal weighting. On CIFAR-10, SVHN, and STL-10, the paper reports consistent gains in accuracy and F1 over five DP-FL baselines under a unified privacy setting. If the claims hold, the framework offers a practical recipe for deploying private federated learning on heterogeneous edge devices with less utility loss.

What carries the argument

Three modules carry the argument. LLDM: a learnable projection W_c ∈ R^(C'×C) with C'≪C reduces the channel dimension of intermediate features, and a magnitude threshold τ_t sparsifies the reduced gradients, producing more structured updates. ADPC: after each round the server sets the next clipping bound C^(t+1) = max(median(S^(t)), C_min) from the norms of client updates, balancing over- and under-clipping. CRA: the server builds nominal weights α_i from exponential moving averages of local validation F1 and update norms, defines an ambiguity set P_t around α with per-client deviation radii and a robustness budget Γ_t, then performs a one-step projected descent to get robust weights p̃ that

What would settle it

A concrete check: compute the ℓ2 sensitivity of the clipped batch-gradient query in Eq. (5) with respect to adding or removing one training sample. If the sensitivity exceeds C(t) — which it does for batch gradients with batch size > 1 — then the Gaussian mechanism in Eq. (12) does not provide (ε,δ)-DP at the claimed ε, and the reported 'under differential privacy' label collapses. A second check: compare the privacy accounting with and without the server's adaptive threshold (Eq. 11) and local validation F1 (Eq. 13) to see whether those statistics leak.

Watch

Extended reading notes

Core claim

FedDimDP's central claim is that the standard DP-FL recipe of fixed gradient clipping plus naive averaging is the root of instability under non-IID data, and that replacing each stage with an adaptive counterpart fixes it. The paper reports that median-based adaptive clipping, local representation compression with gradient sparsification, and ambiguity-set-constrained reweighting of client updates jointly outperform five DP-FL baselines on CIFAR-10, SVHN, and STL-10, with ablation experiments showing each component contributes and robust aggregation mattering most on the harder datasets.

Load-bearing premise

The entire privacy guarantee rests on the assumption that clipping each client's batch gradient to norm C(t) and adding Gaussian noise of variance σ²C(t)² bounds per-sample sensitivity in the standard DP-SGD sense, and that the server's use of update norms and validation F1 scores to adapt thresholds and weights incurs no additional privacy cost — neither of which is proven in the paper.

Editorial extensions

If this is right

Under non-IID data and a fixed privacy budget, adaptive clipping that follows the median update scale avoids both signal loss from over-clipping and noise domination from under-clipping, stabilizing training.
Compressing intermediate features before backpropagation reduces gradient variance, which reduces how much DP noise is amplified during local optimization.
Reweighting client updates by smoothed utility-stability scores and directional deviation keeps the global model closer to a stable descent direction, mitigating client drift.
The framework improves both accuracy and F1 on three image benchmarks relative to five DP-FL baselines, suggesting the three components are complementary rather than redundant.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The method could be tested under stricter privacy budgets (ε<8): the claim that the median-based threshold adapts well while preserving DP is more plausible at moderate ε; at low ε the noise may dominate the median itself, making the threshold a noisy function of private data.
The paper's aggregation uses local validation F1 as a utility signal; a natural extension would be to formalize the privacy cost of releasing these scores, or replace them with a differentially private estimate, to make the end-to-end guarantee rigorous.
The same three-module recipe — representation compression, adaptive clipping, robust reweighting — may transfer to other DP training regimes such as per-sample clipping DP-SGD or decentralized FL, where update norms are also non-stationary.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Short version: this is a clearly written engineering paper whose headline privacy claim does not survive contact with the mechanism. The assembly is new; the privacy proof is not there.

What's genuinely useful: the three modules are known separately, but this specific combination—lightweight channel projection plus gradient sparsification, median-based adaptive clipping, and ambiguity-set robust aggregation driven by EMA-smoothed utility and stability—is new, and the evaluation on CIFAR-10/SVHN/STL-10 is internally consistent. The ablations show each component moves the numbers, the related work is honestly cited, and Algorithm 1 matches the prose.

The soft spot is structural, not cosmetic. Eq. (5) clips the norm of the parameter gradients computed on a whole batch, not per-sample gradients. Replacing one training example can change that clipped batch gradient by more than C(t), so the Gaussian noise in Eq. (12) has no finite sensitivity to be calibrated against. The statement in Section III-B-3 that sensitivity is 'effectively bounded' is unsupported. On top of that, the server's adaptive threshold C(t) is a function of the private updates, and the validation F1 scores that drive aggregation are private data; neither appears in any privacy accountant, RDP or otherwise, and there is no conversion from noise scale to (epsilon, delta). This is not circular reasoning—it is data-dependent adaptivity that needs to be accounted for and is not. The experiments are single runs, with no code, seeds, or standard deviations; Table I is a reasonable sanity check but not evidence of a privacy–utility tradeoff.

If the authors fix the mechanism—per-sample clipping, or a proven bound on the uploaded update, plus an accountant covering threshold adaptation and validation scores—there may be a decent DP-FL paper here, especially the robust aggregation component. As written, the central claim 'under differential privacy' is not established, and I would not cite it as a valid private method.

Recommendation: send it to a DP-literate referee rather than desk-reject out of hand, but expect rejection or major revision. The core ideas are worth engaging; the privacy claim is not.

Referee Report

3 major / 3 minor

Summary. The paper proposes FedDimDP, a federated learning framework that combines three components: a lightweight local dimensionality reduction module (LLDM), an adaptive differentially private gradient clipping mechanism (ADPC), and a constraint-aware robust aggregation scheme (CRA). The authors claim that FedDimDP consistently improves convergence stability and classification performance under differential privacy on CIFAR-10, SVHN, and STL-10 in non-IID settings. The method follows a round-based protocol: clients clip their local gradients, add Gaussian noise, perform local optimization, upload the accumulated model change and a local validation F1 score, and the server adapts the clipping threshold and aggregation weights based on the uploaded statistics. The paper reports experimental results and ablations in Tables I and II.

Significance. If the claimed differential privacy guarantee were valid, the paper would address a practically important problem: improving DP-FL utility under non-IID data and heterogeneous clients. The component-level ablations are informative, and the empirical comparisons span three datasets. However, the central claim of the paper is that the method works 'under differential privacy,' and this claim is not supported by any valid privacy analysis. The algorithm as described releases raw validation F1 scores and clips batch-level gradients rather than per-sample gradients, so the Gaussian mechanism is not calibrated to a finite sensitivity. The adaptive threshold and multi-round composition are also unaccounted for. The paper does not supply a derivation, a privacy accountant, or the parameters needed to verify the claimed ε=8 budget. Because the headline contribution is the privacy-preserving nature of the method, the absence of a sound DP guarantee is a load-bearing flaw that undermines the validity of the reported results.

major comments (3)

[§III-B, Eq. (5) and Eq. (12)] Equation (5) clips the global norm of the client's parameter gradients after backpropagation, i.e., a batch-level (or parameter-tensor-level) gradient, not per-sample gradients. For the Gaussian mechanism to provide differential privacy, the sensitivity of the released quantity with respect to changing one training example must be bounded. Replacing one example in a batch can change the clipped batch gradient by an amount that is not bounded by C unless per-sample gradients are first clipped. Thus the noise N(0, σ²C²I) in Eq. (12) is not calibrated to a valid sensitivity bound. The statement in Section III-B-3 that 'the sensitivity of each client gradient is effectively bounded' is unsupported.
[§III-B-2, Eqs. (10)–(11), and Algorithm 1] The server adapts the clipping threshold C^{(t+1)} from the norms s_i^{(t)} of locally trained updates, which are data-dependent quantities, and this threshold then controls the noise variance in Eq. (12). No adaptive composition or privacy accountant is provided to account for the release and use of C^{(t+1)} across T rounds, nor for the fact that local optimization in Eq. (7) involves E repeated noisy gradient steps. The paper reports only 'ε=8' in Section III-B-3 and never specifies δ, σ, or how ε is derived from the noise scale. Without a composition analysis, the claimed privacy budget is unverifiable.
[§III-C, Eqs. (13)–(16), and Algorithm 1 lines 11–12] Clients upload their local validation F1 score q_i^{(t)} in plaintext, and this score is used to compute the aggregation weights α_i^{(t)} and hence the final global model. A differential privacy mechanism cannot release arbitrary data-dependent statistics without adding noise or proving a sensitivity bound. The validation score is itself a function of private client data; protecting only the model updates while publishing q_i^{(t)} breaks the privacy guarantee. No privacy analysis is given for q_i^{(t)}.

minor comments (3)

[§V, Table I and Table II] The experiments report a single run for each method and dataset, with no standard deviations or confidence intervals. Given the paper's emphasis on 'convergence stability,' error bars are important. Please also state the number of communication rounds, δ, σ, and baseline privacy parameters used in the 'unified differential privacy settings.'
[Algorithm 1] The 'Require' line includes a privacy budget (ε, δ), but the procedure never uses it; no Laplace or Gaussian mechanism parameter is derived from ε in the pseudocode. This makes the algorithm hard to reproduce.
[§III-A and Fig. 2] The sparsification threshold τ_t in Eq. (3) is described as decaying dynamically during training, but no formula or schedule is given. Since sparsification modifies gradients based on their magnitudes, its dependence on private data should also be addressed if the method is to be considered DP.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the DP-guarantee gap is a correctness issue, not a circularity.

full rationale

FedDimDP does not present a formal derivation; its contributions are algorithmic (LLDM, ADPC, CRA) and evaluated empirically against external baselines on CIFAR-10, SVHN, and STL-10. No equation in the paper is solved by substituting back a quantity it was meant to predict. Eq. (11) sets the next clipping threshold from observed update norms, but this is a feedback rule rather than a fitted parameter renamed as a prediction; Table I reports held-out test performance, not the validation scores used by CRA in Eqs. (13)-(16). The references that may share authors with the present paper (e.g., [1], possibly [20]/[29]) are used as background or baselines and are not load-bearing: the central comparison with DP-FedSAM, DP-ACDN, FedACG, AWDP-FL, and FedSA does not reduce to a self-citation. There is a serious correctness gap in the DP claim: Eq. (5) clips the aggregated parameter gradient rather than per-sample gradients, Eq. (12)'s Gaussian noise is calibrated to a data-dependent C(t), and no privacy accountant or composition analysis is supplied. But that is an unsupported privacy guarantee, not circular reasoning. The framework is self-contained against external benchmarks, so the circularity score is 0.

Assumptions & free parameters 7 free parameters · 4 assumptions · 0 invented entities

The paper provides no derivation; it relies on standard DP machinery without verifying sensitivity, and leaves most hyperparameters unspecified. The central privacy guarantee rests on an unproven—and likely false—assumption about batch-gradient clipping.

free parameters (7)

Initial clipping threshold C^(0) and floor C_min
Used in Eq. (11); controls clipping and noise scale, but values are not reported.
Noise scale σ
Claimed to correspond to ε=8, but no composition analysis or numerical value is given.
Sparsification threshold schedule τ_t
Eq. (3) says large early and small later, but no schedule or values are provided.
EMA smoothing factor β
Eqs. (13)–(14); not specified.
Ambiguity radius ρ and robustness budget Γ_t
Eqs. (18)–(19); not specified; they control aggregation reweighting.
Projection step size η_p
Eq. (22); not specified.
Privacy budget ε=8 = 8
Chosen by hand as 'lightweight' with no accounting for rounds, clients, or composition.

assumptions (4)

standard math Gaussian mechanism yields (ε,δ)-DP when noise variance is at least (2 ln(1.25/δ) Δ²)/ε².
Invoked in Sec. III-B3, but the sensitivity Δ is not bounded by batch-gradient clipping.
domain assumption Per-update clipping in federated learning bounds user-level sensitivity.
Cites [24]; the paper assumes it holds for updates derived from noisy local optimization (Eqs. 8, 10).
ad hoc to paper Adaptive composition allows thresholds from private outputs without extra privacy cost.
Not proven; the server updates C(t) via Eq. (11) and uses validation scores q_i in aggregation.
ad hoc to paper Batch-gradient norm clipping bounds per-sample sensitivity.
False; Eq. (5) clips the aggregate gradient, so sensitivity remains unbounded under change of one sample.

how reviews work

0 comments

Cite this review

Pith. "Pith review of An Adaptive Differentially Private Federated Learning Framework." pith.science (2026). https://pith.science/paper/D3RW5MYY

@misc{pith2026260206838,
  author       = {Pith},
  title        = {Pith review of: An Adaptive Differentially Private Federated Learning Framework},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/D3RW5MYY}},
  note         = {Machine review of arXiv:2602.06838}
}

read the original abstract

Federated learning enables collaborative model training across distributed clients while preserving data privacy. However, in practical deployments, device heterogeneity and non-independent and identically distributed (Non-IID) data often lead to unstable and biased gradient. When differential privacy is enforced, conventional fixed gradient clipping and Gaussian noise injection may further amplify gradient perturbations, resulting in training oscillation and degraded model performance. To address these challenges, we propose an adaptive differentially private federated learning framework that explicitly targets model efficiency under heterogeneous and privacy-constrained settings. On the client side, a lightweight local dimensionality reduction module is introduced to learn reduced-dimensional intermediate representations and produce more structured gradients during backpropagation, thereby mitigating noise amplification during local optimization. On the server side, an adaptive gradient clipping strategy dynamically adjusts clipping thresholds based on historical update statistics to avoid over-clipping and noise domination. Furthermore, a constraint-aware robust aggregation mechanism is designed to suppress unreliable or noise-dominated client updates and stabilize global optimization. Extensive experiments on CIFAR-10, SVHN, and STL-10 demonstrate that the proposed method consistently improves convergence stability and classification performance under differential privacy.

Figures

Figures reproduced from arXiv: 2602.06838 by the authors.

**Figure 2.** Overview of the proposed lightweight local dimensionality reduction [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

32 extracted references

[1]

Cellular traffic prediction via byzantine- robust asynchronous federated learning,

H. Ma, K. Yang, and Y . Jiao, “Cellular traffic prediction via byzantine- robust asynchronous federated learning,”IEEE Transactions on Net- work Science and Engineering, vol. 12, pp. 2402–2414, 2025

2025
[2]

Image classification using federated learning,

H. Byeon, A. Ullah, Z. Syedet al., “Image classification using federated learning,” inRecent Trends in Image Processing and Pattern Recognition, 2024, vol. 1839, pp. 66–75

2024
[3]

Enhancing medical image classification via federated learning and pre-trained model,

P. N. Srinivasu, G. J. Lakshmi, S. C. Narahariet al., “Enhancing medical image classification via federated learning and pre-trained model,”The Egyptian Informatics Journal, vol. 27, 2024

2024
[4]

FGS-FL: Enhancing federated learning with differential privacy via flat gradient stream,

J. Hu and H. Zhang, “FGS-FL: Enhancing federated learning with differential privacy via flat gradient stream,”Expert Systems with Applications, vol. 288, 2025

2025
[5]

Federated semi-supervised medical image classification via inter-client relation matching,

Q. Liu, H. Yang, Q. Douet al., “Federated semi-supervised medical image classification via inter-client relation matching,” inMedical Image Computing and Computer-Assisted Intervention, vol. 12905, 2021, pp. 325–335

2021
[6]

Uldp-FL: Federated learning with across-silo user-level differential privacy,

F. Kato, L. Xiong, S. Takagiet al., “Uldp-FL: Federated learning with across-silo user-level differential privacy,”Proceedings of the VLDB Endowment, vol. 17, no. 11, pp. 2826–2839, 2024

2024
[7]

Dyn-D2P: Dynamic decentralized differential privacy for peer-to-peer federated learning,

Y . Rong, K. Duan, S. Heet al., “Dyn-D2P: Dynamic decentralized differential privacy for peer-to-peer federated learning,” inProceedings of the 34th International Joint Conference on Artificial Intelligence, 2025, pp. 7272–7281

2025
[8]

Toward the flatter landscape and better generalization in federated learning under client-level differ- ential privacy,

Y . Shi, K. Wei, L. Shenet al., “Toward the flatter landscape and better generalization in federated learning under client-level differ- ential privacy,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 47, no. 12, pp. 11 632–11 643, 2025

2025

Show all 32 references

[9]

One-shot empirical privacy estimation for federated learning,

G. Andrew, P. Kairouz, S. Ohet al., “One-shot empirical privacy estimation for federated learning,” inProceedings of the International Conference on Learning Representations, 2024

2024
[10]

Asynchronous decentralized federated anomaly detection for 6G networks,

Y . Liu and K. Yang, “Asynchronous decentralized federated anomaly detection for 6G networks,”IEEE Transactions on Cognitive Commu- nications and Networking, 2025

2025
[11]

Belt and braces: When federated learning meets differential privacy,

G. Kamath, A. T. Suresh, and A. Thakurta, “Belt and braces: When federated learning meets differential privacy,”Communications of the ACM, vol. 67, no. 12, pp. 66–77, 2024

2024
[12]

Siren ++: Robust federated learning with proactive alarming and differential privacy,

H. Guo, H. Wang, T. Songet al., “Siren ++: Robust federated learning with proactive alarming and differential privacy,”IEEE Transactions on Dependable and Secure Computing, pp. 5650–5666, 2024

2024
[13]

CAMEL: Communication-efficient and maliciously secure federated learning in the shuffle model of differential privacy,

S. Xu, Y . Zheng, and Z. Hua, “CAMEL: Communication-efficient and maliciously secure federated learning in the shuffle model of differential privacy,” inProceedings of the ACM SIGSAC Conference on Computer and Communications Security, 2024, pp. 243–257

2024
[14]

Noise-aware algorithm for heterogeneous differentially private federated learning,

S. Malekmohammadi, Y . Yu, and Y . Cao, “Noise-aware algorithm for heterogeneous differentially private federated learning,” inProceed- ings of the 41st International Conference on Machine Learning, 2024, pp. 32 738–32 757

2024
[15]

Federated learning with sparsified model perturbation: Improving accuracy under client-level differential privacy,

R. Hu, Y . Gong, and Y . Guo, “Federated learning with sparsified model perturbation: Improving accuracy under client-level differential privacy,”IEEE Transactions on Mobile Computing, vol. 23, no. 8, pp. 8242–8255, 2024

2024
[16]

Harnessing sparsification in federated learning: A secure, efficient, and differentially private realization,

S. Xu, Y . Zheng, and Z. Hua, “Harnessing sparsification in federated learning: A secure, efficient, and differentially private realization,” in Proceedings of the 2025 ACM SIGSAC Conference on Computer and Communications Security, 2025, pp. 2354–2368

2025
[17]

Efficient federated learning privacy preservation method with heterogeneous differential privacy,

J. Ling, J. Zheng, and J. Chen, “Efficient federated learning privacy preservation method with heterogeneous differential privacy,”Comput- ers & Security, vol. 139, 2024

2024
[18]

Differentially private fed- erated learning with time-adaptive privacy spending,

S. Kiani, N. Kulkarni, A. Dziedzicet al., “Differentially private fed- erated learning with time-adaptive privacy spending,” inProceedings of the International Conference on Learning Representations, 2025

2025
[19]

Privacy-preserving federated learning for industrial edge computing via hybrid differential privacy and adaptive compression,

B. Jiang, J. Li, H. Wanget al., “Privacy-preserving federated learning for industrial edge computing via hybrid differential privacy and adaptive compression,”IEEE Transactions on Industrial Informatics, vol. 19, pp. 1136–1144, 2023

2023
[20]

Local differential privacy feder- ated learning based on heterogeneous data multi-privacy mechanism,

J. Wang, Z. Zhang, J. Tianet al., “Local differential privacy feder- ated learning based on heterogeneous data multi-privacy mechanism,” Computer Networks, vol. 254, 2024

2024
[21]

DP-PSAC-FL: Differentially private federated learning based on per-sample adaptive clipping and layer- wise gradient perturbation,

X. Xu, L. Ju, S. Huanget al., “DP-PSAC-FL: Differentially private federated learning based on per-sample adaptive clipping and layer- wise gradient perturbation,”Computer Networks, vol. 261, 2025

2025
[22]

Towards the robustness of dif- ferentially private federated learning,

T. Qi, H. Wang, and Y . Huang, “Towards the robustness of dif- ferentially private federated learning,” inProceedings of the AAAI Conference on Artificial Intelligence, 2024

2024
[23]

Dynamic sparse training with structured sparsity,

M. Lasby, A. Golubeva, U. Evciet al., “Dynamic sparse training with structured sparsity,” inProceedings of the International Conference on Learning Representations, 2024

2024
[24]

An improved analysis of per- sample and per-update clipping in federated learning,

B. Li, X. Jiang, M. N. Schmidtet al., “An improved analysis of per- sample and per-update clipping in federated learning,” inThe Twelfth International Conference on Learning Representations, 2024

2024
[25]

Adaptive clipping for differential private federated learning in interpolation regimes,

T. Fukami, T. Murata, and K. Niwa, “Adaptive clipping for differential private federated learning in interpolation regimes,”Transactions on Machine Learning Research, 2025

2025
[26]

Exponential moving average of weights in deep learning: Dynamics and benefits,

D. Morales-Brotons, T. V ogels, and H. Hendrikx, “Exponential moving average of weights in deep learning: Dynamics and benefits,”Trans- actions on Machine Learning Research, 2024

2024
[27]

Distributed distributionally robust optimization with non-convex objectives,

Y . Jiao, K. Yang, and D. Song, “Distributed distributionally robust optimization with non-convex objectives,” inAdvances in Neural Information Processing Systems, vol. 35, 2022, pp. 7987–7999

2022
[28]

Understanding global ag- gregation and optimization of federated learning,

S. I. Nanayakkara, S. R. Pokhrel, and G. Li, “Understanding global ag- gregation and optimization of federated learning,”Future Generation Computer Systems, vol. 159, pp. 114–133, 2024

2024
[29]

Self-adaptive asynchronous federated optimizer with adversarial sharpness-aware minimization,

X. Zhang, J. Wang, W. Baoet al., “Self-adaptive asynchronous federated optimizer with adversarial sharpness-aware minimization,” Future Generation Computer Systems, vol. 161, pp. 638–654, 2024

2024
[30]

Differentially private federated learning with an adaptive noise mechanism,

R. Xue, K. Xue, B. Zhuet al., “Differentially private federated learning with an adaptive noise mechanism,”IEEE Transactions on Information Forensics and Security, vol. 19, pp. 74–87, 2024

2024
[31]

Differentially private federated learning on Non-IID data: Convergence analysis and adaptive opti- mization,

L. Chen, X. Ding, Z. Baoet al., “Differentially private federated learning on Non-IID data: Convergence analysis and adaptive opti- mization,”IEEE Transactions on Knowledge and Data Engineering, vol. 36, no. 9, pp. 4567–4581, 2024

2024
[32]

A lightweight differentially private federated learning framework for edge intelligence,

Y . Zhang, X. Chen, Z. Liet al., “A lightweight differentially private federated learning framework for edge intelligence,”IEEE Internet of Things Journal, vol. 11, no. 7, pp. 12 045–12 058, 2024

2024

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.

[1] [1]

Cellular traffic prediction via byzantine- robust asynchronous federated learning,

H. Ma, K. Yang, and Y . Jiao, “Cellular traffic prediction via byzantine- robust asynchronous federated learning,”IEEE Transactions on Net- work Science and Engineering, vol. 12, pp. 2402–2414, 2025

2025

[2] [2]

Image classification using federated learning,

H. Byeon, A. Ullah, Z. Syedet al., “Image classification using federated learning,” inRecent Trends in Image Processing and Pattern Recognition, 2024, vol. 1839, pp. 66–75

2024

[3] [3]

Enhancing medical image classification via federated learning and pre-trained model,

P. N. Srinivasu, G. J. Lakshmi, S. C. Narahariet al., “Enhancing medical image classification via federated learning and pre-trained model,”The Egyptian Informatics Journal, vol. 27, 2024

2024

[4] [4]

FGS-FL: Enhancing federated learning with differential privacy via flat gradient stream,

J. Hu and H. Zhang, “FGS-FL: Enhancing federated learning with differential privacy via flat gradient stream,”Expert Systems with Applications, vol. 288, 2025

2025

[5] [5]

Federated semi-supervised medical image classification via inter-client relation matching,

Q. Liu, H. Yang, Q. Douet al., “Federated semi-supervised medical image classification via inter-client relation matching,” inMedical Image Computing and Computer-Assisted Intervention, vol. 12905, 2021, pp. 325–335

2021

[6] [6]

Uldp-FL: Federated learning with across-silo user-level differential privacy,

F. Kato, L. Xiong, S. Takagiet al., “Uldp-FL: Federated learning with across-silo user-level differential privacy,”Proceedings of the VLDB Endowment, vol. 17, no. 11, pp. 2826–2839, 2024

2024

[7] [7]

Dyn-D2P: Dynamic decentralized differential privacy for peer-to-peer federated learning,

Y . Rong, K. Duan, S. Heet al., “Dyn-D2P: Dynamic decentralized differential privacy for peer-to-peer federated learning,” inProceedings of the 34th International Joint Conference on Artificial Intelligence, 2025, pp. 7272–7281

2025

[8] [8]

Toward the flatter landscape and better generalization in federated learning under client-level differ- ential privacy,

Y . Shi, K. Wei, L. Shenet al., “Toward the flatter landscape and better generalization in federated learning under client-level differ- ential privacy,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 47, no. 12, pp. 11 632–11 643, 2025

2025