Pith. sign in

REVIEW 4 major objections 5 minor 83 references

FedRP claims that sending only low-dimensional random projections of model weights can keep federated accuracy near FedAvg while adding a formal differential-privacy guarantee and shrinking communication cost.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review

2026-08-04 18:15 UTC pith:EM2LYL5W

load-bearing objection FedRP's core idea is interesting, but the DP proof and the projected-consensus bridge don't hold up as written; with corrections it could be worth another look. the 4 major comments →

arxiv 2509.10041 v1 pith:EM2LYL5W submitted 2025-09-12 cs.LG

FedRP: A Communication-Efficient Approach for Differentially Private Federated Learning Using Random Projection

Mohammad Hasan Narimani , Mostafa Tavassolipour This is my paper
classification cs.LG
keywords federated learningdifferential privacyrandom projectionADMMconsensus optimizationcommunication efficiencyreconstruction attackdimensionality reduction
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

FedRP tries to solve two federated-learning pains at once: the cost of sending full model updates and the privacy risk of sending gradients that can be reverse-engineered. The algorithm has each client multiply its local model by a shared random matrix and send only the low-dimensional projected vector to the server; the server averages these vectors, and ADMM drives the clients toward agreement in the projected space. The paper's central theoretical claim is that this scheme is (epsilon, delta)-differentially private with epsilon = (Delta/sigma_min)(m + sqrt(8m ln(1/delta))) per round, where m is the projected dimension. Experiments on MNIST and CIFAR-10/100 with LeNet-5, ResNet-18, and VGG16 indicate accuracy close to standard FedAvg, better than FedAvg with added DP noise, and communication reduced to a few bytes per round when m=1. If these claims hold, privacy and bandwidth no longer have to be traded against accuracy in cross-silo federated learning.

Core claim

For each communication round, clients generate a common random projection matrix A (m x n, entries i.i.d. N(0,1/n)), project local weights w_i to z_i = A w_i, send z_i to the server, and receive the average bar-z. The local update objective is augmented with the term (rho/2)||A w_i - bar-z||^2 so that ADMM pushes the projections to consensus. The paper proves Theorem 3.1: the mechanism is (epsilon, delta)-DP with epsilon = (Delta/sigma_min)(m + sqrt(8m ln(1/delta))), using the fact that z_i is Gaussian with covariance scaled by ||w_i||^2 and bounding the ratio of densities of z and z' for neighboring datasets. It then asserts that because the projected parameters are aligned, the original pa

What carries the argument

The random projection matrix A, shared among clients but hidden from the server, is the load-bearing object. It simultaneously compresses the model from n to m dimensions (the communication saving), hides the full model from the server and attackers (the privacy mechanism), and defines the consensus target in the ADMM loss. The DP proof treats the projected vector z = A w as a Gaussian random variable whose covariance is determined by ||w||^2, and uses the ratio of densities between neighboring models to bound the privacy loss.

Load-bearing premise

The claim that agreement among clients in the low-dimensional projected space implies the high-dimensional local models are similar enough that one client's full model can be returned as the global model; with m=1 many different full models share the same projected value.

What would settle it

Train any of the reported setups with m=1 and record the accuracy of the single client model chosen as the global model, then compare it with the accuracy of the FedAvg global model; if the chosen model lags FedAvg by a large margin while the projected values agree, the consensus-in-projection assumption is what failed. More directly, the assumption is false whenever two valid solutions w1 != w2 satisfy A w1 = A w2 for the same random A, which for m=1 happens for any difference vector in the null space of A.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X Bluesky LinkedIn Reddit HN

If this is right

  • Per client per round, communication drops from n parameters to m scalars; with m=1 that is a few bytes regardless of model size.
  • The per-round DP guarantee is quantified; total privacy loss over T rounds accumulates roughly linearly as T*epsilon under composition.
  • Because full gradients and model snapshots never reach the server, common reconstruction attacks that need those signals lose their input.
  • Reduced projection dimension improves the stated privacy bound; the paper reports accuracy remains near FedAvg down to m=1 in its experiments.
  • Noise-free projection avoids the accuracy collapse that the paper observes when DP noise is added to pre-trained models such as VGG16.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The biggest unproven step is the inference from consensus in the projected space to consensus in the full parameter space; with m=1, any two models differing only in the null space of A produce identical projected values, so the server cannot certify that the picked client's model represents the group.
  • A direct test of that step would be to record the full-model distance between clients under FedRP and compare it with FedAvg; if FedRP's client models remain far apart while their projections agree, the final-model selection is not justified.
  • The current analysis assumes all clients participate every round (cross-silo); extending the same projection trick to cross-device FL with partial participation and dropouts is an open direction the paper explicitly leaves out.
  • The per-round DP proof bounds the ratio of densities for a single projected message; the server actually sees an average of projected models, so the exact privacy guarantee of the aggregated output is left at the level of composition rather than a direct sensitivity analysis of the average.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes FedRP, a federated learning algorithm that combines random projection with ADMM. Each client transmits a low-dimensional projection z_i = A w_i of its model parameters to the server, which averages the projections and returns the aggregate. The authors claim that this scheme yields a strong (epsilon, delta)-differential privacy guarantee (Theorem 3.1), reduces communication cost, and achieves accuracy close to FedAvg. Experiments compare FedRP to FedAvg, FedADMM, FedAvg+DP, and FWC on MNIST, CIFAR-10, and CIFAR-100 with LeNet-5, ResNet-18, and VGG16. The paper also makes a communication-efficiency claim in Table 2 and includes a proof of the privacy theorem in Appendix A.

Significance. If the central claims were correct, FedRP would be an interesting contribution: it promises privacy without adding noise, strong communication savings, and accuracy competitive with FedAvg. The paper is also commendable for providing code and for testing across several model architectures. However, the load-bearing elements of the paper are not supported by the current analysis. The privacy theorem uses a different distribution for the projection matrix than the algorithm, the privacy bound depends on an unreported and uncontrolled parameter sigma_min, and the utility/communication argument relies on an unproven bridge from consensus in the projected space to consensus in the full parameter space. These issues affect the core claims, not just the presentation.

major comments (4)
  1. [Section 3.2 vs. Appendix A] The algorithm in Section 3.2 generates A with entries drawn from N(0,1/n), while Appendix A states 'A is a random matrix whose elements are drawn from standard normal distribution.' Consequently, the distributional claim in the proof, z ~ N(0, ||w||_2^2 I_m), is incorrect for the actual mechanism. With entries of variance 1/n, the covariance is (||w||_2^2 / n) I_m, which changes the density ratio and the resulting epsilon expression in Eq. (15). The proof must be reworked to match the algorithm's scaling or the algorithm must use a standard normal matrix.
  2. [Theorem 3.1 and Table 3] Eq. (15) expresses epsilon in terms of Delta and sigma_min, but sigma_min is not a parameter the algorithm enforces or reports. The algorithm does not clip model parameters, does not bound their L2 norm away from zero, and does not compute or enforce the claimed sensitivity Delta. As a result, the numbers labeled epsilon/Delta in Table 3 cannot be reproduced from the manuscript and appear to be controlled by an unreported sigma_min. Since sigma_min can be chosen to make epsilon arbitrarily small, the reported privacy guarantee is partly a restatement of an assumed constant rather than a property of the mechanism. The authors need to specify exactly how sigma_min and Delta are computed, or provide a mechanism that enforces them.
  3. [Section 3.2, paragraph 5] The claim 'Since the projected parameters are aligned, it is highly likely that the original model parameters w_i are also similar across clients' is load-bearing and unsupported. The ADMM objective in Eq. (9) penalizes only A w_i - zbar; the null space of A is unpenalized. With m=1 and T rounds, the transmitted projections impose at most T linear constraints on each n-dimensional w_i, leaving a large subspace in which client models can differ while their projections agree. If the server selects a single client's model, there is no guarantee it is near consensus; if the server averages full client models, it must receive the full models, which contradicts the communication savings in Table 2. A formal convergence or concentration argument is needed for the null-space components, and no such argument appears in the paper.
  4. [Section 4.3.1 and Algorithm 1] The test-phase description is ambiguous and directly affects the utility/communication claim. The paper states that 'the central server can use either a randomly selected client model or the average of all client models,' but Algorithm 1 transmits only z_i, not the full model w_i. If the reported accuracies use the average of all full client models, then the server must receive all full models, invalidating the communication-cost comparison. If they use a randomly selected client model, the accuracy results depend on the unproven full-space consensus assumption. The experiments should clarify which option was used and report communication costs accordingly.
minor comments (5)
  1. [Section 3.2 and Appendix A] Notation is inconsistent: Section 3.2 defines A as m x n and z = A w, while Appendix A uses A in R^{n x m} and z = A^T w. This should be harmonized.
  2. [Section 3.3, Definition 2] Definition 2 says epsilon and delta are in [0,1], but Theorem 3.1 later allows delta in (0,1). The ranges should be stated consistently.
  3. [Algorithm 1 and Section 3.2] Algorithm 1 says 'Randomly generate A^{t+1}' but the text describes a secure seed-sharing scheme via the server. The pseudocode should reflect the intended protocol for generating and distributing A.
  4. [Section 4.5] The resistance-to-reconstruction-attack claim is not supported by experiments or a formal argument. The statement that 'there is no analytical and error-free solution' does not rule out approximate reconstruction attacks; this should be substantiated or softened.
  5. [Table 2] The table uses the abbreviation 'RPD' without defining it, and the units 'Byte' vs 'KB' should be checked for consistency (e.g., RPD=1000 for LeNet-5 is listed as 4 KB, which is plausible, but the header formatting is unclear).

Circularity Check

1 steps flagged

Privacy guarantee reduces to an unreported assumed constant sigma_min; the reported epsilon values are computed from this input, making the central privacy claim partly circular.

specific steps
  1. fitted input called prediction [Theorem 3.1, Section 3.3 (Eq. 15), Table 3]
    "Theorem 3.1.[( ϵ, δ)-DP Privacy of the FedRP Algorithm] The FedRP algorithm is(ϵ, δ)-DP with ϵ= ∆ σmin ( m+ q 8mln 1 δ ! ) (15) for any value ofδ∈ (0, 1), where∆is the L2 sensitivity, m is the reduced-dimensional model parameters obtained through random projection andσmin is the minimum allowable L2 norm of the vector of model parametersw."

    The claimed privacy parameter ε is defined in Eq. 15 as a direct function of σ_min, a constant that is not measured, reported, or enforced by Algorithm 1. The paper states 'a lower bound σ_min is imposed,' but no mechanism imposes it. Table 3's ε/Δ values (e.g., 1.92, 2.35 for RPD=1) are evaluated from Eq. 15 with an unstated σ_min; choosing a larger σ_min makes ε arbitrarily small. Thus the reported privacy 'prediction' is a restatement of the assumed lower bound on ||w||, not an independently measured or benchmarked property of the algorithm. The guarantee reduces by construction to an input assumption.

full rationale

The paper's main theoretical contribution is the DP guarantee in Theorem 3.1. That guarantee is conditional on two assumptions: an L2 sensitivity bound Δ and a lower bound σ_min on the norm of every client model. The proof derives the formula from these assumptions. However, the algorithm does not actually enforce σ_min (no clipping or normalization is described), and the paper never reports the value of σ_min used to produce Table 3. Consequently, the privacy numbers are not a measured or externally fixed property of FedRP; they are a direct evaluation of a formula whose key input is an unreported free parameter. This is a fitted-input-called-prediction pattern: the ε values in Table 3 are forced by the chosen σ_min, so the central privacy claim is partly circular. The other major weakness—the assertion in Section 3.2 that equality of projected parameters implies similarity of full model parameters—is an unsupported heuristic, not a circular reduction. The test-phase statement allowing the server to use the average of all client models also suggests the accuracy results may depend on full-model transmission, undermining the communication-efficiency claim, but this is a correctness/experimental-design issue rather than circularity. No load-bearing self-citations or imported uniqueness theorems were found; the ADMM and JL-lemma citations are standard external results. Overall, the paper is not fully circular, but the headline privacy guarantee is substantially weakened by its dependence on an unverified, unreported input constant.

Axiom & Free-Parameter Ledger

4 free parameters · 6 axioms · 0 invented entities

The central claims rest on the secret shared projection matrix, the projection-to-full-consensus inference, and an assumed norm lower bound sigma_min that is not enforced or reported. These are domain assumptions or ad hoc assumptions rather than derived facts.

free parameters (4)
  • sigma_min (minimum L2 norm of model parameters) = not reported
    Introduced in the proof of Theorem 3.1 to prevent w from being zero; the epsilon bound is inversely proportional to it. Table 3 reports epsilon/Delta values that depend on undisclosed per-model sigma_min, so the privacy guarantee is not reproducible.
  • Delta (L2 sensitivity) = not reported
    Assumed in Theorem 3.1 and Table 3; no clipping or sensitivity computation is provided, so epsilon/Delta cannot be converted to an actual epsilon guarantee.
  • rho (ADMM penalty parameter) = not reported
    Appears in the augmented Lagrangian and the client update in equation 10; its value is not stated in the experiments.
  • Random projection dimension m = 1, 10, 50, 100, 1000, or 10000 depending on model
    This hyperparameter directly controls communication cost and the claimed epsilon, and it is tuned per experiment in Table 1.
axioms (6)
  • domain assumption Client data are identically distributed and IID across clients.
    Section 3.1 states 'we assume that the distribution of client data is identical and that all samples within the clients are distributed independently and identically (IID).' This is a strong restriction relative to much FL literature.
  • domain assumption All clients participate in every training round.
    Section 1 explicitly narrows the scope to the cross-silo setting where all clients participate; the algorithm and the average in equation 13 depend on this.
  • domain assumption Clients can securely share a random projection matrix (seed) without revealing it to the server.
    Section 3.2 relies on public-key encryption of the seed relayed through the server; no attack model or implementation details are given.
  • ad hoc to paper Closeness of projected vectors A w_i implies closeness of full models w_i.
    Section 3.2 states: 'Since the projected parameters are aligned, it is highly likely that the original model parameters w_i are also similar across clients.' This is unproved and central to the utility claim.
  • ad hoc to paper Model parameters w and w' have L2 norm at least sigma_min.
    Appendix A assumes ||w||_2 >= sigma_min and ||w'||_2 >= sigma_min; the algorithm does not enforce this bound, and the DP guarantee disappears when w approaches zero.
  • standard math Standard chi-square tail bounds and DP composition theorems.
    Used in Appendix A for Lemma 2 and in Section 3.3 to claim the total privacy budget accumulates roughly as T times epsilon over T rounds.

reviewed 2026-08-04 · how reviews work

0 comments
Cite this review

Pith. "Pith review of FedRP: A Communication-Efficient Approach for Differentially Private Federated Learning Using Random Projection." pith.science (2026). https://pith.science/paper/EM2LYL5W

@misc{pith2026250910041,
  author       = {Pith},
  title        = {Pith review of: FedRP: A Communication-Efficient Approach for Differentially Private Federated Learning Using Random Projection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EM2LYL5W}},
  note         = {Machine review of arXiv:2509.10041}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Federated learning (FL) offers an innovative paradigm for collaborative model training across decentralized devices, such as smartphones, balancing enhanced predictive performance with the protection of user privacy in sensitive areas like Internet of Things (IoT) and medical data analysis. Despite its advantages, FL encounters significant challenges related to user privacy protection against potential attacks and the management of communication costs. This paper introduces a novel federated learning algorithm called FedRP, which integrates random projection techniques with the Alternating Direction Method of Multipliers (ADMM) optimization framework. This approach enhances privacy by employing random projection to reduce the dimensionality of model parameters prior to their transmission to a central server, reducing the communication cost. The proposed algorithm offers a strong $(\epsilon, \delta)$-differential privacy guarantee, demonstrating resilience against data reconstruction attacks. Experimental results reveal that FedRP not only maintains high model accuracy but also outperforms existing methods, including conventional differential privacy approaches and FedADMM, in terms of both privacy preservation and communication efficiency.

Figures

Figures reproduced from arXiv: 2509.10041 by Mohammad Hasan Narimani, Mostafa Tavassolipour.

Figure 1
Figure 1. Figure 1: Structure of the FedRP. In each round t, each client first updates its local model w t+1 i . Then, a new random projection matrix At+1 is generated and multiplied by w t+1 i to compute the compressed vector z t+1 i . These vectors are sent to the server, which aggregates them as z¯ t+1 and returns the result to the clients. This process repeats until convergence. is then applied to the model parameters of … view at source ↗
Figure 2
Figure 2. Figure 2: Comparison of FedRP Algorithm Performance with Other Algorithms Under [PITH_FULL_IMAGE:figures/full_fig_p018_2.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

83 extracted references · 6 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in ":" * " " * FUNCTION f...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in ":" * " " * FUNCTION f...

  3. [3]

    , author Muntakim, A

    author Ahmed, M. , author Muntakim, A. , author Tabassum, N. , author Rahim, M.A. , author Shah, F.M. , year 2024 . title On-device federated learning in smartphones for detecting depression from reddit posts . https://arxiv.org/abs/2410.13709, http://arxiv.org/abs/2410.13709 arXiv:2410.13709

    Pith/arXiv arXiv 2024

  4. [4]

    , author Moustafa, N

    author Akter, M. , author Moustafa, N. , year 2024 . title Federated learning-based privacy protection methods for internet of things systems . :10.5121/csit.2024.141103

    arXiv 2024

  5. [5]

    , author Mahdavifar, H

    author Aldaghri, N. , author Mahdavifar, H. , author Beirami, A. , year 2023 . title Federated learning with heterogeneous differential privacy . https://arxiv.org/abs/2110.15252, http://arxiv.org/abs/2110.15252 arXiv:2110.15252

    Pith/arXiv arXiv 2023

  6. [6]

    , author Shaukat, S

    author Asad, M. , author Shaukat, S. , author Hu, D. , author Wang, Z. , author Javanmardi, E. , author Nakazato, J. , author Tsukada, M. , year 2023 . title Limitations and future aspects of communication costs in federated learning: A survey . journal Sensors (Basel, Switzerland) volume 23 . https://api.semanticscholar.org/CorpusID:261165129

    2023

  7. [7]

    , author Duchi, J

    author Bhowmick, A. , author Duchi, J. , author Freudiger, J. , author Kapoor, G. , author Rogers, R. , year 2019 . title Protection against reconstruction and its applications in private federated learning . https://arxiv.org/abs/1812.00984, http://arxiv.org/abs/1812.00984 arXiv:1812.00984

    Pith/arXiv arXiv 2019

  8. [8]

    , author Dras, M

    author Biswas, S. , author Dras, M. , author Faustini, P. , author Fernandes, N. , author McIver, A. , author Palamidessi, C. , author Sadeghi, P. , year 2024 . title Bayes' capacity as a measure for reconstruction attacks in federated learning . https://arxiv.org/abs/2406.13569, http://arxiv.org/abs/2406.13569 arXiv:2406.13569

    Pith/arXiv arXiv 2024

  9. [9]

    , author Parikh, N

    author Boyd, S. , author Parikh, N. , author Chu, E. , author Peleato, B. , author Eckstein, J. , year 2011 . title Distributed optimization and statistical learning via the alternating direction method of multipliers . journal Found. Trends Mach. Learn. volume 3 , pages 1–122 . https://doi.org/10.1561/2200000016, :10.1561/2200000016

    doi:10.1561/2200000016 2011

  10. [10]

    , author Gao, D

    author Chen, D. , author Gao, D. , author Xie, Y. , author Pan, X. , author Li, Z. , author Li, Y. , author Ding, B. , author Zhou, J. , year 2023 a. title Fs-real: Towards real-world cross-device federated learning . https://arxiv.org/abs/2303.13363, http://arxiv.org/abs/2303.13363 arXiv:2303.13363

    Pith/arXiv arXiv 2023

  11. [11]

    , author Yan, H

    author Chen, J. , author Yan, H. , author Liu, Z. , author Zhang, M. , author Xiong, H. , author Yu, S. , year 2024 . title When federated learning meets privacy-preserving computation . journal ACM Comput. Surv. volume 56 . https://doi.org/10.1145/3679013, :10.1145/3679013

    work page doi:10.1145/3679013 2024

  12. [12]

    , author Choquette-Choo, C.A

    author Chen, W.N. , author Choquette-Choo, C.A. , author Kairouz, P. , author Suresh, A.T. , year 2022 . title The fundamental price of secure aggregation in differentially private federated learning . https://arxiv.org/abs/2203.03761, http://arxiv.org/abs/2203.03761 arXiv:2203.03761

    Pith/arXiv arXiv 2022

  13. [13]

    , author Song, D

    author Chen, W.N. , author Song, D. , author Ozgur, A. , author Kairouz, P. , year 2023 b. title Privacy amplification via compression: Achieving the optimal privacy-accuracy-communication trade-off in distributed mean estimation . https://arxiv.org/abs/2304.01541, http://arxiv.org/abs/2304.01541 arXiv:2304.01541

    Pith/arXiv arXiv 2023

  14. [14]

    , author Schneider, T

    author Demmler, D. , author Schneider, T. , author Zohner, M. , year 2015 . title Aby - a framework for efficient mixed-protocol secure two-party computation , in: booktitle Network and Distributed System Security Symposium . https://api.semanticscholar.org/CorpusID:14639818

    2015

  15. [15]

    , year 2012

    author Deng, L. , year 2012 . title The mnist database of handwritten digit images for machine learning research . journal IEEE Signal Processing Magazine volume 29 , pages 141--142

    2012

  16. [16]

    , author Shirke, P

    author Dhade, P. , author Shirke, P. , year 2023 . title Federated learning for healthcare: A comprehensive review . journal Engineering Proceedings volume 59 . https://www.mdpi.com/2673-4591/59/1/230, :10.3390/engproc2023059230

    work page doi:10.3390/engproc2023059230 2023

  17. [17]

    , author McSherry, F

    author Dwork, C. , author McSherry, F. , author Nissim, K. , author Smith, A. , year 2006 . title Calibrating noise to sensitivity in private data analysis , in: editor Halevi, S. , editor Rabin, T. (Eds.), booktitle Theory of Cryptography , publisher Springer Berlin Heidelberg , address Berlin, Heidelberg . pp. pages 265--284

    2006

  18. [18]

    , author Roth, A

    author Dwork, C. , author Roth, A. , year 2014 . title The algorithmic foundations of differential privacy . journal Found. Trends Theor. Comput. Sci. volume 9 , pages 211–407 . https://doi.org/10.1561/0400000042, :10.1561/0400000042

    doi:10.1561/0400000042 2014

  19. [19]

    , year 1985

    author ElGamal, T. , year 1985 . title A public key cryptosystem and a signature scheme based on discrete logarithms , in: editor Blakley, G.R. , editor Chaum, D. (Eds.), booktitle Advances in Cryptology , publisher Springer Berlin Heidelberg , address Berlin, Heidelberg . pp. pages 10--18

    1985

  20. [20]

    , author Arief, B

    author ElZemity, A. , author Arief, B. , year 2024 . title Privacy threats and countermeasures in federated learning for internet of things: A systematic review . https://arxiv.org/abs/2407.18096, http://arxiv.org/abs/2407.18096 arXiv:2407.18096

    Pith/arXiv arXiv 2024

  21. [21]

    , year 2021

    author Freksen, C.B. , year 2021 . title An introduction to johnson-lindenstrauss transforms . https://arxiv.org/abs/2103.00564, http://arxiv.org/abs/2103.00564 arXiv:2103.00564

    Pith/arXiv arXiv 2021

  22. [22]

    , author Chen, H

    author Fu, C. , author Chen, H. , author Ruan, N. , year 2025 . title Privacy for free: Spy attack in vertical federated learning by both active and passive parties . journal IEEE Transactions on Information Forensics and Security volume 20 , pages 2550--2563 . :10.1109/TIFS.2025.3534469

    arXiv 2025

  23. [23]

    , author Chen, Z

    author Fu, J. , author Chen, Z. , author Han, X. , year 2022 . title Adap dp-fl: Differentially private federated learning with adaptive noise . https://arxiv.org/abs/2211.15893, http://arxiv.org/abs/2211.15893 arXiv:2211.15893

    Pith/arXiv arXiv 2022

  24. [24]

    , year 2009

    author Gentry, C. , year 2009 . title Fully homomorphic encryption using ideal lattices , in: booktitle Proceedings of the Forty-First Annual ACM Symposium on Theory of Computing , publisher Association for Computing Machinery , address New York, NY, USA . p. pages 169–178 . https://doi.org/10.1145/1536414.1536440, :10.1145/1536414.1536440

    arXiv 2009

  25. [25]

    , author Klein, T

    author Geyer, R.C. , author Klein, T. , author Nabi, M. , year 2017 . title Differentially private federated learning: A client level perspective . journal CoRR volume abs/1712.07557 . http://arxiv.org/abs/1712.07557, http://arxiv.org/abs/1712.07557 arXiv:1712.07557

    Pith/arXiv arXiv 2017

  26. [26]

    , author Pan, K

    author Gong, M. , author Pan, K. , author Xie, Y. , author Qin, A. , author Tang, Z. , year 2020 . title Preserving differential privacy in deep neural networks with relevance-based adaptive noise imposition . journal Neural Networks volume 125 . :10.1016/j.neunet.2020.02.001

    work page doi:10.1016/j.neunet.2020.02.001 2020

  27. [27]

    , author Li, Y

    author Gong, Y. , author Li, Y. , author Freris, N.M. , year 2022 . title Fedadmm: A robust federated deep learning framework with adaptivity to system heterogeneity . https://arxiv.org/abs/2204.03529, http://arxiv.org/abs/2204.03529 arXiv:2204.03529

    Pith/arXiv arXiv 2022

  28. [28]

    , author Shirazi, H

    author Gorbett, M. , author Shirazi, H. , author Ray, I. , year 2024 . title Cross-silo federated learning across divergent domains with iterative parameter alignment . https://arxiv.org/abs/2311.04818, http://arxiv.org/abs/2311.04818 arXiv:2311.04818

    Pith/arXiv arXiv 2024

  29. [29]

    , author Yap, P.T

    author Guan, H. , author Yap, P.T. , author Bozoki, A. , author Liu, M. , year 2024 . title Federated learning for medical image analysis: A survey . https://arxiv.org/abs/2306.05980, http://arxiv.org/abs/2306.05980 arXiv:2306.05980

    Pith/arXiv arXiv 2024

  30. [30]

    , author Karimi, B

    author Haddadpour, F. , author Karimi, B. , author Li, P. , author Li, X. , year 2020 . title Fedsketch: Communication-efficient and private federated learning via sketching . https://arxiv.org/abs/2008.04975, http://arxiv.org/abs/2008.04975 arXiv:2008.04975

    Pith/arXiv arXiv 2020

  31. [31]

    , author Rao, K

    author Hard, A. , author Rao, K. , author Mathews, R. , author Ramaswamy, S. , author Beaufays, F. , author Augenstein, S. , author Eichner, H. , author Kiddon, C. , author Ramage, D. , year 2019 . title Federated learning for mobile keyboard prediction . https://arxiv.org/abs/1811.03604, http://arxiv.org/abs/1811.03604 arXiv:1811.03604

    Pith/arXiv arXiv 2019

  32. [32]

    O zkasap, \

    author Hayyolalam, V. , author \"O zkasap, \"O . , year 2023 . title Communication efficient client selection in federated learning . https://api.semanticscholar.org/CorpusID:260140691

    2023

  33. [33]

    , author Zhang, X

    author He, K. , author Zhang, X. , author Ren, S. , author Sun, J. , year 2015 . title Deep residual learning for image recognition . https://arxiv.org/abs/1512.03385, http://arxiv.org/abs/1512.03385 arXiv:1512.03385

    Pith/arXiv arXiv 2015

  34. [34]

    , author Huang, J

    author Huang, C. , author Huang, J. , author Liu, X. , year 2022 . title Cross-silo federated learning: Challenges and opportunities . https://arxiv.org/abs/2206.12949, http://arxiv.org/abs/2206.12949 arXiv:2206.12949

    Pith/arXiv arXiv 2022

  35. [35]

    , author Lindenstrauss, J

    author Johnson, W. , author Lindenstrauss, J. , year 1984 . title Extensions of lipschitz maps into a hilbert space . journal Contemporary Mathematics volume 26 , pages 189--206 . :10.1090/conm/026/737400

    doi:10.1090/conm/026/737400 1984

  36. [36]

    , author McMahan, H.B

    author Kairouz, P. , author McMahan, H.B. , author Avent, B. , author Bellet, A. , author Bennis, M. , author Bhagoji, A.N. , author Bonawitz, K.A. , author Charles, Z. , author Cormode, G. , author Cummings, R. , author D'Oliveira, R.G.L. , author Rouayheb, S.E. , author Evans, D. , author Gardner, J. , author Garrett, Z. , author Gasc \' o n, A. , autho...

    Pith/arXiv arXiv 2019

  37. [37]

    , year 2009

    author Krizhevsky, A. , year 2009 . title Learning multiple layers of features from tiny images , pages 32--33 https://www.cs.toronto.edu/ kriz/learning-features-2009-TR.pdf

    2009

  38. [38]

    , author Bottou, L

    author LeCun, Y. , author Bottou, L. , author Bengio, Y. , author Haffner, P. , year 1998 . title Gradient-based learning applied to document recognition . journal Proceedings of the IEEE volume 86 , pages 2278--2323 . :10.1109/5.726791

    doi:10.1109/5.726791 1998

  39. [39]

    , author Khodak, M

    author Li, J. , author Khodak, M. , author Caldas, S. , author Talwalkar, A. , year 2020 . title Differentially private meta-learning . https://arxiv.org/abs/1909.05830, http://arxiv.org/abs/1909.05830 arXiv:1909.05830

    Pith/arXiv arXiv 2020

  40. [40]

    , author Wen, Z

    author Li, Q. , author Wen, Z. , author Wu, Z. , author Hu, S. , author Wang, N. , author Li, Y. , author Liu, X. , author He, B. , year 2023 . title A survey on federated learning systems: Vision, hype and reality for data privacy and protection . journal IEEE Transactions on Knowledge and Data Engineering volume 35 , pages 3347--3366 . :10.1109/TKDE.202...

    arXiv 2023

  41. [41]

    , author Weeraddana, P.C

    author Magnússon, S. , author Weeraddana, P.C. , author Rabbat, M.G. , author Fischione, C. , year 2015 . title On the convergence of alternating direction lagrangian methods for nonconvex structured optimization problems . https://arxiv.org/abs/1409.8033, http://arxiv.org/abs/1409.8033 arXiv:1409.8033

    Pith/arXiv arXiv 2015

  42. [42]

    , author Ramage, D

    author McMahan, B. , author Ramage, D. , year 2017 . title Federated learning: Collaborative machine learning without centralized training data . journal Google Research Blog volume 3

    2017

  43. [43]

    , author Andrew, G

    author McMahan, H.B. , author Andrew, G. , author Erlingsson, U. , author Chien, S. , author Mironov, I. , author Papernot, N. , author Kairouz, P. , year 2019 . title A general approach to adding differential privacy to iterative training procedures . https://arxiv.org/abs/1812.06210, http://arxiv.org/abs/1812.06210 arXiv:1812.06210

    Pith/arXiv arXiv 2019

  44. [44]

    , author Moore, E

    author McMahan, H.B. , author Moore, E. , author Ramage, D. , author Hampson, S. , author y Arcas, B.A. , year 2023 . title Communication-efficient learning of deep networks from decentralized data . https://arxiv.org/abs/1602.05629, http://arxiv.org/abs/1602.05629 arXiv:1602.05629

    Pith/arXiv arXiv 2023

  45. [45]

    , author Yang, T

    author Mendieta, M. , author Yang, T. , author Wang, P. , author Lee, M. , author Ding, Z. , author Chen, C. , year 2021 . title Local learning matters: Rethinking data heterogeneity in federated learning . journal CoRR volume abs/2111.14213 . https://arxiv.org/abs/2111.14213, http://arxiv.org/abs/2111.14213 arXiv:2111.14213

    Pith/arXiv arXiv 2021

  46. [46]

    , year 2017

    author Nabil, M. , year 2017 . title Random projection and its applications . https://arxiv.org/abs/1710.03163, http://arxiv.org/abs/1710.03163 arXiv:1710.03163

    Pith/arXiv arXiv 2017

  47. [47]

    , author Abdelhadi, A

    author Ouadrhiri, A.E. , author Abdelhadi, A. , year 2022 . title Differential privacy for deep and federated learning: A survey . journal IEEE Access volume 10 , pages 22359--22380 . :10.1109/ACCESS.2022.3151670

    arXiv 2022

  48. [48]

    , year 1999

    author Paillier, P. , year 1999 . title Public-key cryptosystems based on composite degree residuosity classes , in: editor Stern, J. (Ed.), booktitle Advances in Cryptology --- EUROCRYPT '99 , publisher Springer Berlin Heidelberg , address Berlin, Heidelberg . pp. pages 223--238

    1999

  49. [49]

    , author Hytla, P

    author PHAN, D.N. , author Hytla, P. , author Rice, A. , author Nguyen, N. , year 2025 . title Federated learning with randomized alternating direction method of multipliers and application in training neural networks . journal Neural Networks volume 189 , pages 107501 . :10.1016/j.neunet.2025.107501

    arXiv 2025

  50. [50]

    , author Aono, Y

    author Phong, L.T. , author Aono, Y. , author Hayashi, T. , author Wang, L. , author Moriai, S. , year 2018 . title Privacy-preserving deep learning via additively homomorphic encryption . journal IEEE Transactions on Information Forensics and Security volume 13 , pages 1333--1345 . :10.1109/TIFS.2017.2787987

    arXiv 2018

  51. [51]

    , author Mukhtar Dirir, A

    author Habib ur Rehman, M. , author Mukhtar Dirir, A. , author Salah, K. , author Svetinovic, D. , year 2020 . title Fairfed: Cross-device fair federated learning , in: booktitle 2020 IEEE Applied Imagery Pattern Recognition Workshop (AIPR) , pp. pages 1--7 . :10.1109/AIPR50011.2020.9425266

    arXiv 2020

  52. [52]

    , author Panda, A

    author Rothchild, D. , author Panda, A. , author Ullah, E. , author Ivkin, N. , author Stoica, I. , author Braverman, V. , author Gonzalez, J. , author Arora, R. , year 2020 . title Fetchsgd: Communication-efficient federated learning with sketching . https://arxiv.org/abs/2007.07682, http://arxiv.org/abs/2007.07682 arXiv:2007.07682

    Pith/arXiv arXiv 2020

  53. [53]

    , author Edwards, B

    author Sheller, M. , author Edwards, B. , author Reina, G. , author Martin, J. , author Pati, S. , author Kotrotsou, A. , author Milchenko, M. , author Xu, W. , author Marcus, D. , author Colen, R. , author Bakas, S. , year 2020 . title Federated learning in medicine: facilitating multi-institutional collaborations without sharing patient data . journal S...

    doi:10.1038/s41598-020-69250-1 2020

  54. [54]

    , author Zisserman, A

    author Simonyan, K. , author Zisserman, A. , year 2015 . title Very deep convolutional networks for large-scale image recognition . https://arxiv.org/abs/1409.1556, http://arxiv.org/abs/1409.1556 arXiv:1409.1556

    Pith/arXiv arXiv 2015

  55. [55]

    , author Wang, Z

    author Song, Y. , author Wang, Z. , author Zuazua, E. , year 2024 . title Fedadmm-insa: An inexact and self-adaptive admm for federated learning . https://arxiv.org/abs/2402.13989, http://arxiv.org/abs/2402.13989 arXiv:2402.13989

    Pith/arXiv arXiv 2024

  56. [56]

    , author Yu, Z

    author Song, Z. , author Yu, Z. , author Zhang, L. , year 2022 . title Iterative sketching and its application to federated learning . https://openreview.net/forum?id=U_Jog0t3fAu

    2022

  57. [57]

    , author Karamzade, A

    author Tavassolipour, M. , author Karamzade, A. , author Mirzaeifard, R. , author Motahari, S.A. , author Shalmani, M.T.M. , year 2018 . title Structure learning of sparse ggms over multiple access networks . https://arxiv.org/abs/1812.10437, http://arxiv.org/abs/1812.10437 arXiv:1812.10437

    Pith/arXiv arXiv 2018

  58. [58]

    , author Motahari, S.A

    author Tavassolipour, M. , author Motahari, S.A. , author Shalmani, M.T.M. , year 2017 . title Learning of gaussian processes in distributed and communication limited systems . https://arxiv.org/abs/1705.02627, http://arxiv.org/abs/1705.02627 arXiv:1705.02627

    Pith/arXiv arXiv 2017

  59. [59]

    , author Motahari, S.A

    author Tavassolipour, M. , author Motahari, S.A. , author Shalmani, M.T.M. , year 2019 . title Learning of tree-structured gaussian graphical models on distributed data under communication constraints . journal IEEE Transactions on Signal Processing volume 67 , pages 17–28 . http://dx.doi.org/10.1109/TSP.2018.2876325, :10.1109/tsp.2018.2876325

    arXiv 2019

  60. [60]

    , author AL-Jumaili, A.S.A

    author Tayyeh, H.K. , author AL-Jumaili, A.S.A. , year 2024 . title Balancing privacy and performance: A differential privacy approach in federated learning . journal Computers volume 13 . https://www.mdpi.com/2073-431X/13/11/277, :10.3390/computers13110277

    work page doi:10.3390/computers13110277 2024

  61. [61]

    , author Jin, L

    author Teo, Z. , author Jin, L. , author Liu, N. , author Li, S. , author Miao, D. , author Zhang, X. , author Ng, W. , author Tan, T. , author Lee, D. , author Chua, K. , author Heng, J. , author Liu, Y. , author Goh, R. , author Ting, D. , year 2024 . title Federated machine learning in healthcare: A systematic review on clinical applications and techni...

    arXiv 2024

  62. [62]

    , author Baracaldo, N

    author Truex, S. , author Baracaldo, N. , author Anwar, A. , author Steinke, T. , author Ludwig, H. , author Zhang, R. , year 2018 . title A hybrid approach to privacy-preserving federated learning . journal CoRR volume abs/1812.03224 . http://arxiv.org/abs/1812.03224, http://arxiv.org/abs/1812.03224 arXiv:1812.03224

    Pith/arXiv arXiv 2018

  63. [63]

    , author Zhang, H

    author Wang, K. , author Zhang, H. , author Kaddoum, G. , author Shin, H. , author Quek, T.Q.S. , author Win, M.Z. , year 2025 . title Sgan-ra: Reconstruction attack for big model in asynchronous federated learning . journal IEEE Communications Magazine volume 63 , pages 66--72 . :10.1109/MCOM.001.2400062

    work page doi:10.1109/mcom.001.2400062 2025

  64. [64]

    , author Zhu, T

    author Wang, L. , author Zhu, T. , author Zhou, W. , author Yu, P.S. , year 2024 . title Linkage on security, privacy and fairness in federated learning: New balances and new perspectives . https://arxiv.org/abs/2406.10884, http://arxiv.org/abs/2406.10884 arXiv:2406.10884

    Pith/arXiv arXiv 2024

  65. [65]

    , author Yin, W

    author Wang, Y. , author Yin, W. , author Zeng, J. , year 2018 . title Global convergence of admm in nonconvex nonsmooth optimization . https://arxiv.org/abs/1511.06324, http://arxiv.org/abs/1511.06324 arXiv:1511.06324

    Pith/arXiv arXiv 2018

  66. [66]

    , author Li, J

    author Wei, K. , author Li, J. , author Ding, M. , author Ma, C. , author Yang, H.H. , author Farokhi, F. , author Jin, S. , author Quek, T.Q. , author Poor, H.V. , year 2020 a. title Federated learning with differential privacy: Algorithms and performance analysis . journal IEEE transactions on information forensics and security volume 15 , pages 3454--3469

    2020

  67. [67]

    , author Li, J

    author Wei, K. , author Li, J. , author Ding, M. , author Ma, C. , author Yang, H.H. , author Farokhi, F. , author Jin, S. , author Quek, T.Q.S. , author Vincent Poor, H. , year 2020 b. title Federated learning with differential privacy: Algorithms and performance analysis . journal IEEE Transactions on Information Forensics and Security volume 15 , pages...

    arXiv 2020

  68. [68]

    , author Zhao, S

    author Wei, Y. , author Zhao, S. , author Zhao, C. , author Liu, Z. , author Chen, Z. , author Zhao, M. , year 2025 . title Fedcaprivacy: Privacy-preserving heterogeneous federated learning with anonymous adaptive clustering . https://arxiv.org/abs/2503.23292, http://arxiv.org/abs/2503.23292 arXiv:2503.23292

    Pith/arXiv arXiv 2025

  69. [69]

    , author Zhang, Z

    author Wen, J. , author Zhang, Z. , author Lan, Y. , author sheng Cui, Z. , author Cai, J. , author Zhang, W. , year 2022 . title A survey on federated learning: challenges and applications . journal International Journal of Machine Learning and Cybernetics volume 14 , pages 513 -- 535 . https://api.semanticscholar.org/CorpusID:253496346

    2022

  70. [70]

    , author Han, J

    author Xia, T. , author Han, J. , author Ghosh, A. , author Mascolo, C. , year 2023 . title Cross-device federated learning for mobile health diagnostics: A first study on covid-19 detection . https://arxiv.org/abs/2303.07067, http://arxiv.org/abs/2303.07067 arXiv:2303.07067

    Pith/arXiv arXiv 2023

  71. [71]

    , author Wen, Q

    author Xu, B. , author Wen, Q. , author Cheng, L. , author Hu, X. , author Li, T. , author Sun, Y. , year 2025 . title Gradient inversion attack via image-correction-penalty based over-parameterized regression network in federated learning . journal IEEE Internet of Things Journal , pages 1--1 :10.1109/JIOT.2025.3572806

    arXiv 2025

  72. [72]

    , author Noura, H.N

    author Yaacoub, J.P.A. , author Noura, H.N. , author Salman, O. , year 2023 . title Security of federated learning with iot systems: Issues, limitations, challenges, and solutions . journal Internet of Things and Cyber-Physical Systems volume 3 , pages 155--179 . https://www.sciencedirect.com/science/article/pii/S2667345223000226, :https://doi.org/10.1016...

    work page doi:10.1016/j.iotcps.2023.04.001 2023

  73. [73]

    , author Wang, Q

    author Yang, C. , author Wang, Q. , author Xu, M. , author Wang, S. , author Bian, K. , author Liu, X. , year 2020 a. title Heterogeneity-aware federated learning . journal CoRR volume abs/2006.06983 . https://arxiv.org/abs/2006.06983, http://arxiv.org/abs/2006.06983 arXiv:2006.06983

    Pith/arXiv arXiv 2020

  74. [74]

    , author Lyu, L

    author Yang, M. , author Lyu, L. , author Zhao, J. , author Zhu, T. , author Lam, K.Y. , year 2020 b. title Local differential privacy and its applications: A comprehensive survey . https://arxiv.org/abs/2008.03686, http://arxiv.org/abs/2008.03686 arXiv:2008.03686

    Pith/arXiv arXiv 2020

  75. [75]

    , author Jia, Q.S

    author Yang, Y. , author Jia, Q.S. , author Xu, Z. , author Guan, X. , author Spanos, C.J. , year 2022 . title Proximal admm for nonconvex and nonsmooth optimization . journal Automatica volume 146 , pages 110551 . https://www.sciencedirect.com/science/article/pii/S0005109822004125, :https://doi.org/10.1016/j.automatica.2022.110551

    arXiv 2022

  76. [76]

    , year 1982

    author Yao, A.C. , year 1982 . title Protocols for secure computations , in: booktitle 23rd annual symposium on foundations of computer science (sfcs 1982) , organization IEEE . pp. pages 160--164

    1982

  77. [77]

    , author Mallya, A

    author Yin, H. , author Mallya, A. , author Vahdat, A. , author Alvarez, J.M. , author Kautz, J. , author Molchanov, P. , year 2021 . title See through gradients: Image batch recovery via gradinversion . https://arxiv.org/abs/2104.07586, http://arxiv.org/abs/2104.07586 arXiv:2104.07586

    Pith/arXiv arXiv 2021

  78. [78]

    , author Dong, X

    author You, Z. , author Dong, X. , author Li, S. , author Liu, X. , author Ma, S. , author Shen, Y. , year 2025 . title Local differential privacy is not enough: A sample reconstruction attack against federated learning with local differential privacy . journal IEEE Transactions on Information Forensics and Security volume 20 , pages 1519--1534 . :10.1109...

    arXiv 2025

  79. [79]

    , year 2024

    author Yuan, G. , year 2024 . title Admm for nonconvex optimization under minimal continuity assumption . https://arxiv.org/abs/2405.03233, http://arxiv.org/abs/2405.03233 arXiv:2405.03233

    Pith/arXiv arXiv 2024

  80. [80]

    , author Barakat, S

    author Zerka, F. , author Barakat, S. , author Walsh, S. , author Bogowicz, M. , author Leijenaar, R.T.H. , author Jochems, A. , author Miraglio, B. , author Townend, D. , author Lambin, P. , year 2020 . title Systematic review of privacy-preserving distributed machine learning from federated databases in health care . journal JCO Clinical Cancer Informat...

    doi:10.1200/cci.19.00047 2020

Showing first 80 references.

This paper was first reviewed by deepseek-v4-flash on August 4, 2026.