Pith. sign in

REVIEW 3 major objections 5 minor 33 references

Bayesian Robust Aggregation for Federated Learning

T0 review · 3 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

Pith's one-line read A Bayesian aggregation rule infers which federated-learning clients are honest, removing the need to specify the number of malicious clients.

desk verdict The stated Algorithm 1 never leaves the all-benign fixed point, so the paper's central robustness claim is unsupported as written. read the letter →

arxiv 2505.02490 v1 pith:56TXS7RG submitted 2025-05-05 cs.LG stat.ML

classification cs.LGstat.ML
keywords federatedlearningrobustaggregationByzantineattacksBayesianinferencevariationalempiricalBayesbackdoorattackadversarialrobustness
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 an aggregation rule for federated learning in which every client update is assigned a posterior probability of being honest, and the global model is the weighted mean of the updates with those probabilities as weights. The key move is to marginalize over Bernoulli latent variables indicating honesty, through an evidence lower bound, so the method infers the fraction of malicious clients at each communication round instead of requiring it as input. The claim is that this Bayesian rule is as effective as Krum against strong attacks such as sign flipping and backdoors, while remaining as simple as the sample mean and adapting when attackers appear, disappear, or change behavior. A careful reader would care because existing robust aggregators either need a user-specified number of compromised clients or degrade on complex tasks exactly when the adversarial share is high and variable.

What carries the argument

The central object is the evidence lower bound of Equation (11): a Gaussian likelihood over the squared residual norms $\lVert w_k - w\rVert_2^2 / \sigma^2$, latent Bernoulli variables $b_k$ for client honesty, and a KL-divergence term whose prior $\varepsilon = M/K$ is itself optimized, making the procedure an empirical-Bayes estimate. The machinery is coordinate ascent: update per-client posterior probabilities $\pi_k$ with Equation (15), then re-estimate $w$ and $\sigma^2$ in closed form from Equations (14). This converts a combinatorial subset-selection problem, Equation (5), into $O(TK)$ iterations, and lets the posterior probabilities perform outlier rejection automatically, so the method adapts to a static or changing number of malicious clients without a user-supplied count.

What would settle it

Construct a communication round with $K=20$ fixed update vectors in which $M=4$ malicious vectors lie just inside the convex hull of the honest ones, run Algorithm 1, and check whether the learned probabilities $\pi_k$ assign them near-zero weight and whether the output $w$ matches the honest mean. If the ELBO optimum is pulled toward the malicious vectors, then the variational solution is not $(M,\kappa)$-robust even though Proposition 2.1 applies to the combinatorial estimator.

Watch

Extended reading notes

Core claim

The central claim is that maximizing the marginalized likelihood, with each client's honesty represented by a latent Bernoulli variable, defines a robust mean update whose robustness does not depend on knowing the number of compromised clients. Algorithm 1 maximizes the corresponding ELBO by alternating closed-form updates: the posterior probability $\pi_k$ for each client from Equation (15), the mean $w$ as the $\pi$-weighted average of updates, and the scale $\sigma^2$ as the $\pi$-weighted mean squared residual. The paper proves that the hard subset-selection estimator being relaxed satisfies an $(M,\kappa)$-robustness bound (Proposition 2.1), and it reports empirically that the variational solution matches or exceeds Krum's test accuracy and backdoor attack success rate on MNIST, Fashion-MNIST, and CIFAR-10, with malicious fractions up to $45\%$ that may vary across rounds.

Load-bearing premise

The load-bearing premise is that the variational ELBO solution computed by Algorithm 1 inherits the $(M,\kappa)$-robustness that is proved only for the hard subset-selection estimator of Equation (5) in Appendix C; no proof of the transfer is given, and the Gaussian likelihood over squared residual norms is the heuristic that makes the transfer plausible.

Editorial extensions

If this is right

  • On CIFAR-10 under sign flipping and backdoor attacks with 40% malicious clients, the method keeps test accuracy within about 1% of the non-adversarial baseline, a level reached only by Krum among the compared defenses.
  • In all three configurations, the method suppresses backdoor attack success rate to at most 8% on CIFAR-10, while Median, Trimmed Mean, and Geometric Median leave attack success rates near 30–100%.
  • The method requires no hyperparameter for the number of malicious clients, unlike Krum's $L$ and Trimmed Mean's $\beta$, and it tracks a dynamically changing adversarial share by estimating $\varepsilon$ at every round.
  • In a fully benign setting the method's accuracy matches Federated Averaging, so robustness does not come at a measurable utility cost in the tested configurations.
  • Because each round only computes weighted means and scalar updates, the aggregation cost stays $O(TK)$ and remains comparable to coordinate-wise median.

Reading between the lines

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

  • The paper proves the robustness bound only for the hard subset-selection estimator of Equation (5); a targeted test of whether the ELBO optimum still satisfies the bound when malicious updates sit just inside the convex hull of honest updates would settle this implicit transfer step.
  • Because the prior share $\varepsilon$ is re-estimated every round, the method could double as an attack-monitoring tool by reporting the inferred fraction of compromised clients over time, a use the authors do not develop.
  • The scalar-residual Gaussian likelihood throws away per-coordinate information, so a structured extension that respects the KL scale issue noted in Section 4.3 may detect stealthy attacks localized in a few coordinates.
  • An untested stress direction is stronger data heterogeneity, since the experiments use Dirichlet $\alpha = 1.0$ and $0.5$; much smaller values could break the single-cluster assumption about honest updates.
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.

Referee Report

3 major / 5 minor

Summary. The paper proposes a Bayesian robust aggregation rule for federated learning in which the global update is computed as a weighted average of client models, with weights interpreted as posterior probabilities of client honesty. The authors derive the method by first formulating robust aggregation as a hard subset-selection problem (Eq. 5), prove an (M,κ)-robustness bound for that problem (Proposition 2.1), and then relax it into a variational objective (ELBO) over latent Bernoulli indicators of honesty (Eqs. 9-11), optimized by the EM-style Algorithm 1. The paper claims the resulting method is adaptive to an unknown or varying number of malicious clients, requires no user-specified hyperparameter such as the number of adversaries, and matches Krum's accuracy and backdoor-attack resistance on MNIST, Fashion-MNIST, and CIFAR-10. The empirical section compares the method against FEDAVG, Median, Trimmed Mean, Geometric Median, and Krum under four attack types and static/dynamic adversary configurations.

Significance. If the algorithm worked as specified, the paper would make a valuable practical contribution: a Byzantine-robust aggregation rule that avoids the common requirement of knowing the number of compromised clients, has simple weighted-mean updates, is backed by a theoretical robustness bound, and is evaluated on a broad benchmark (three datasets, four attacks, static and dynamic adversaries). The paper also provides public code and an ablation study, and the careful comparison against multiple baselines is a clear strength. However, the significance is heavily contingent on the algorithm actually implementing the described inference; as written, the central algorithm does not do so, which substantially weakens the paper's claims.

major comments (3)
  1. [Section 2, Algorithm 1, Eq. (15)] Algorithm 1 as written is degenerate. It initializes π_k = 1 for all k (line 2), so Σπ_old = K and the factor (K − Σπ_old)/Σπ_old in the π-update of Eq. (15) equals 0 at the first iteration. Consequently, π_new_k = 1 for every k regardless of the likelihood p(w_k|w,σ²). The subsequent updates of w and σ² (lines 7–8) are then exactly the unweighted mean and variance, and the same argument repeats at every iteration: (1,...,1) is a fixed point of the EM map. Thus, Algorithm 1 as written reduces to FEDAVG and cannot adapt to any static or dynamic attack. This contradicts the central robustness claim and the probabilities shown in Figure 1. The empirical results in Tables 2–4 cannot be attributed to the procedure specified in the paper unless the implementation uses an unstated initialization (e.g., π_k < 1 or ε > 0), which would be an essential undocumented component. Please correct the algorithm or explicitly state the actual initialization and update scheme used to generate the reported results.
  2. [Section 2, Eqs. (9)–(11) and Proposition 2.1] The robustness guarantee in Proposition 2.1 is proven for the hard subset-selection estimator in Eq. (5), which enforces a hard cardinality constraint |S| = K − M. The actual algorithm optimizes the ELBO in Eq. (11), a variational relaxation with a soft prior over indicators and an empirically estimated ε given by Eq. (12). No argument is given that the solution of Eq. (11) inherits the (M,κ)-robustness of Proposition 2.1. The sentence following Eq. (13) only motivates the Gaussian likelihood by its appearance in Eq. (5); it does not establish that the variational estimator satisfies an analogous bound. The paper should either prove that the Algorithm 1 output satisfies a similar robustness guarantee with respect to a suitable set of benign clients, or clearly state that Proposition 2.1 applies only to the intermediate combinatorial problem and that the robustness of Algorithm 1 rests solely on the empirical evaluation.
  3. [Appendix C, Eq. (25) vs. main text Eq. (6)] There is a technical inconsistency between the statement of Proposition 2.1 and its proof. The main-text bound in Eq. (6) involves a sum of norms, specifically (κ/(K−M)) Σ_{k∈B} ∥w_k − w_B∥₂, while the appendix proof concludes in Eq. (25) a bound with Σ_{k∈B} ∥w_k − w_B∥₂² (squared norms) and a different constant. In addition, Eq. (20) of the appendix appears to mix squared and unsquared norms. Please reconcile the theorem statement with the proof and verify which bound actually follows from the argument.
minor comments (5)
  1. [Abstract, Section 2, Algorithm 1] The method is described as 'parameter-free' and 'hyperparameter-free', but Algorithm 1 depends on the initialization of π_k and σ², the EM convergence criterion, and the maximum number of iterations T. Please qualify this claim and state the exact values used in the experiments.
  2. [Figure 1] The lower heatmap in Figure 1 shows the estimated benign probabilities varying across clients and rounds, which is inconsistent with the all-ones posterior produced by Algorithm 1 as written. The figure and its caption should be updated to describe what the actual implementation computes.
  3. [Section 2, Eq. (8) to Eq. (9)] The transition from Eq. (8) to Eq. (9) is described as a relaxation, but the expectation under the prior p(b) with ε = M/K is not equivalent to the cardinality-constrained maximum in Eq. (8). Please clarify the precise relationship between the two objectives.
  4. [Section 2, Computational complexity] The claim that Algorithm 1 'converges to a local maximum of the ELBO' is not supported by a derivation of Eq. (15) as a coordinate-ascent step for the ELBO in Eq. (11); the update is simply cited to (Karakulev et al., 2024). Please provide the derivation or a precise reference to the convergence result.
  5. [Abstract, Tables 7–8] The abstract states that the method 'consistently achieves state-of-the-art performance', but the appendix results show cases where the method is below Krum (e.g., FMNIST Random attack at 40% malicious, ACC 0.83 vs. 0.89) and otherwise at parity. Please align the abstract's claim with the reported results.

Circularity Check

1 steps flagged · score 8.0 of 10

Algorithm 1's own initialization makes Eq. (15) the identity map, reducing the claimed robust aggregation to FEDAVG by construction.

  1. other [Algorithm 1 (lines 2 and 6), Equations (12) and (15), Section 2]
    "Algorithm 1 (lines 2 and 6): 'πk = 1 for k = 1,...,K' and 'update π← with Equation (15) using w and σ2'. Eq. (12): 'ε = 1−Σπ_k/K'. Eq. (15): 'πnew_k = [1 + (K−Σπ_old)/(Σπ_old) · 1/p(w_k|w,σ2)]^{-1}'."

    At initialization Σπ_old=K, so the prefactor (K−Σπ_old)/Σπ_old in Eq. (15) is 0, giving π_new=1 for every k regardless of the likelihood. The subsequent w- and σ²-updates (Algorithm 1 lines 7–8) are therefore exactly the unweighted mean and variance; this repeats identically every iteration, so (1,...,1) is a fixed point of the EM map. Consequently the written procedure is FEDAVG and cannot adapt to static or dynamic attacks, contrary to the central claim. The robustness 'prediction' is not derived but is equivalent to the input average by construction; reproducing Tables 2–4 would require an unstated initialization (e.g., ε>0 or π≠1), an essential undocumented component.

full rationale

The paper's central claim is adaptive robustness to a varying number of Byzantine clients. The derivation chain starts from the combinatorial robust estimator Eq. (5) with Proposition 2.1, then relaxes it to the ELBO Eq. (11) and to Algorithm 1. The relaxation itself is a legitimate variational method, and the self-citation to Karakulev et al. (2024) is a published, independent framework rather than an unverified self-supporting theorem, so by itself it does not create circularity. The Proposition 2.1 bound, however, is proved for Eq. (5), not for the ELBO solution; the paper asserts rather than proves that the variational optimum inherits the (M,κ)-robustness, which is a gap but not a circular step. The decisive circular-by-construction issue is in Algorithm 1: with π initialized to 1, the factor (K−Σπ_old)/Σπ_old in Eq. (15) is 0 at every iteration, so π_new=1 for all clients. The weighted updates for w and σ² then coincide exactly with the unweighted sample mean and variance, i.e., FEDAVG. Hence the written algorithm cannot downweight any client, and the reported state-of-the-art accuracy/ASR in Tables 2–4 cannot be produced by the stated procedure; the adaptive robustness claim reduces to plain averaging by construction. An unstated change (e.g., π<1 initialization or a different ε handling) would be required, which is an essential undocumented component. This makes the central derivation self-defeating, warranting a score of 8.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The central claim rests on three main loads: a non-standard Gaussian residual model, a standard mean-field approximation, and an unproven transfer of robustness from a combinatorial estimator to a variational relaxation. The only user-chosen quantities are the initial trust vector and the stopping rule, plus the data-fitted scale σ².

free parameters (3)
  • Initial posterior probabilities π_k = 1 for all k
    Algorithm 1 starts by trusting every client (ε=0), and the final estimate can depend on this starting point; the paper does not analyze sensitivity to this initialization.
  • EM convergence criterion / maximum iterations T = not specified
    Algorithm 1 loops until convergence, but the paper gives no tolerance or maximum iteration count, even though the computational complexity is stated as O(TK).
  • Gaussian variance σ² (per-round scale) = estimated iteratively from data via Eq (14)
    The scale parameter is fit by the EM procedure and directly affects the posterior weights in Eq (15); it is a data-dependent parameter of the aggregation rule.
assumptions (3)
  • domain assumption Benign client updates, after centering, are treated as Gaussian with common mean w and variance σ², applied to the scalar residuals ∥w_k−w∥² in Eq (13).
    This is a contamination-model assumption from robust statistics, but squared Euclidean norms are non-negative and not Gaussian; the paper itself flags the dimensional inconsistency in Section 4.3 and proceeds heuristically.
  • standard math The posterior over latent honesty indicators factorizes across clients (mean-field variational approximation).
    Standard mean-field variational inference (Blei et al., 2017), cited in the paper; the approximation is standard but inexact.
  • ad hoc to paper The solution of the intractable marginalized objective in Eq (9) or the ELBO in Eq (11) is robust in the same sense as the subset-centroid estimator in Eq (5), for which Proposition 2.1 is proven.
    The paper proves robustness only for the hard subset-selection rule in Appendix C, not for the variational EM algorithm actually proposed in Algorithm 1; the transfer is asserted without proof.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Bayesian Robust Aggregation for Federated Learning." pith.science (2026). https://pith.science/paper/56TXS7RG

@misc{pith2026250502490,
  author       = {Pith},
  title        = {Pith review of: Bayesian Robust Aggregation for Federated Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/56TXS7RG}},
  note         = {Machine review of arXiv:2505.02490}
}
read the original abstract

Federated Learning enables collaborative training of machine learning models on decentralized data. This scheme, however, is vulnerable to adversarial attacks, when some of the clients submit corrupted model updates. In real-world scenarios, the total number of compromised clients is typically unknown, with the extent of attacks potentially varying over time. To address these challenges, we propose an adaptive approach for robust aggregation of model updates based on Bayesian inference. The mean update is defined by the maximum of the likelihood marginalized over probabilities of each client to be `honest'. As a result, the method shares the simplicity of the classical average estimators (e.g., sample mean or geometric median), being independent of the number of compromised clients. At the same time, it is as effective against attacks as methods specifically tailored to Federated Learning, such as Krum. We compare our approach with other aggregation schemes in federated setting on three benchmark image classification data sets. The proposed method consistently achieves state-of-the-art performance across various attack types with static and varying number of malicious clients.

Figures

Figures reproduced from arXiv: 2505.02490 by the authors.

Figure 1
Figure 1. Federated Learning for classifying FMNIST data using 25 rounds and 10 clients. Attacks are dynamic: in some commu￾nication rounds, malicious clients submit honest updates (upper heatmap). Our aggregation method estimates probability for each client of being ‘benign’ (lower heatmap), which results in a global model with high test accuracy (bottom). et al., 2021; Cheng et al., 2021; Ye et al., 2020; Jiang et al., 2020… view at source ↗
Figure 2
Figure 2. Ablation study. (a) Test accuracy (ACC) under Sign Flip with fraction of malicious clients varying from 0 to 40% and (b) attack success rate (ASR) under Backdoor attack with fraction of malicious clients varying from 10% to 40% [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Examples of backdoor-triggered samples from (a) MNIST, (b) Fashion-MNIST, and (c) CIFAR-10 datasets. indicate that using only the first step significantly degrades performance on benign tasks, while using only the second step results in a low attack success rate (ASR). Additionally, for the CIFAR-10 dataset, we found that ini￾tiating the attack from the beginning of training prevents the model from converging. There… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Ablation study. Test accuracy (ACC) under (a) Random and (b) Label-flipping attacks, with fraction of malicious clients varying from 0 to 40% [PITH_FULL_IMAGE:figures/full_fig_p013_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 18 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 format.date year duplicate empty "emp...

  2. [2]

    Fixing by mixing: A recipe for optimal byzantine ml under heterogeneity

    Allouah, Y., Farhadkhani, S., Guerraoui, R., Gupta, N., Pinot, R., and Stephan, J. Fixing by mixing: A recipe for optimal byzantine ml under heterogeneity. In International Conference on Artificial Intelligence and Statistics, pp.\ 1232--1300. PMLR, 2023

  3. [3]

    How to backdoor federated learning

    Bagdasaryan, E., Veit, A., Hua, Y., Estrin, D., and Shmatikov, V. How to backdoor federated learning. CoRR, abs/1807.00459, 2018. URL http://arxiv.org/abs/1807.00459

  4. [4]

    M., Guerraoui, R., and Stainer, J

    Blanchard, P., El Mhamdi, E. M., Guerraoui, R., and Stainer, J. Machine Learning with Adversaries : Byzantine Tolerant Gradient Descent . In Advances in Neural Information Processing Systems , volume 30. Curran Associates, Inc., 2017

  5. [5]

    M., Kucukelbir, A., and McAuliffe, J

    Blei, D. M., Kucukelbir, A., and McAuliffe, J. D. Variational inference: A review for statisticians. Journal of the American statistical Association, 112 0 (518): 0 859--877, 2017

  6. [6]

    Secureboost: A lossless federated learning framework

    Cheng, K., Fan, T., Jin, Y., Liu, Y., Chen, T., Papadopoulos, D., and Yang, Q. Secureboost: A lossless federated learning framework. IEEE intelligent systems, 36 0 (6): 0 87--98, 2021

  7. [7]

    R., Zhong, A., Harouni, A., Gentili, A., Abidin, A

    Dayan, I., Roth, H. R., Zhong, A., Harouni, A., Gentili, A., Abidin, A. Z., Liu, A., Costa, A. B., Wood, B. J., Tsai, C.-S., et al. Federated learning for predicting clinical outcomes in patients with covid-19. Nature medicine, 27 0 (10): 0 1735--1743, 2021

  8. [8]

    The mnist database of handwritten digit images for machine learning research [best of the web]

    Deng, L. The mnist database of handwritten digit images for machine learning research [best of the web]. IEEE Signal Processing Magazine, 29 0 (6): 0 141--142, 2012. doi:10.1109/MSP.2012.2211477

Show all 33 references
  1. [9]

    Fang, M., Cao, X., Jia, J., and Gong, N. Z. Local model poisoning attacks to byzantine-robust federated learning. In Proceedings of the 29th USENIX Conference on Security Symposium , SEC '20, pp.\ 1623--1640, USA, August 2020. USENIX Association

  2. [10]

    and Chen, J

    Fang, P. and Chen, J. On the vulnerability of backdoor defenses for federated learning. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, pp.\ 11800--11808, 2023

  3. [11]

    Variance reduction is an antidote to byzantines: Better rates, weaker assumptions and communication compression as a cherry on the top

    Gorbunov, E., Horv \'a th, S., Richt \'a rik, P., and Gidel, G. Variance reduction is an antidote to byzantines: Better rates, weaker assumptions and communication compression as a cherry on the top. In The Eleventh International Conference on Learning Representations

  4. [12]

    Deep residual learning for image recognition

    He, K., Zhang, X., Ren, S., and Sun, J. Deep residual learning for image recognition. CoRR, abs/1512.03385, 2015. URL http://arxiv.org/abs/1512.03385

  5. [13]

    Huber, P. J. Robust statistical procedures. SIAM, 1996

  6. [14]

    Customized federated learning for accelerated edge computing with heterogeneous task targets

    Jiang, H., Liu, M., Yang, B., Liu, Q., Li, J., and Guo, X. Customized federated learning for accelerated edge computing with heterogeneous task targets. Computer Networks, 183: 0 107569, 2020

  7. [15]

    Adaptive robust learning using latent bernoulli variables

    Karakulev, A., Zachariah, D., and Singh, P. Adaptive robust learning using latent bernoulli variables. In Proceedings of the 41st International Conference on Machine Learning, pp.\ 23105--23122, 2024

  8. [16]

    P., He, L., and Jaggi, M

    Karimireddy, S. P., He, L., and Jaggi, M. Byzantine-robust learning on heterogeneous datasets via bucketing. In International Conference on Learning Representations

  9. [17]

    P., He, L., and Jaggi, M

    Karimireddy, S. P., He, L., and Jaggi, M. Learning from History for Byzantine Robust Optimization . In ICML 2021 - Proceedings of International Conference on Machine Learning, 2021. URL https://arxiv.org/abs/2012.10333

  10. [18]

    and Hinton, G

    Krizhevsky, A. and Hinton, G. Learning multiple layers of features from tiny images. Technical Report 0, University of Toronto, Toronto, Ontario, 2009. URL https://www.cs.toronto.edu/ kriz/learning-features-2009-TR.pdf

  11. [19]

    Gradient-based learning applied to document recognition

    Lecun, Y., Bottou, L., Bengio, Y., and Haffner, P. Gradient-based learning applied to document recognition. Proceedings of the IEEE, 86 0 (11): 0 2278--2324, 1998. doi:10.1109/5.726791

  12. [20]

    MacKay, D. J. Information theory, inference and learning algorithms. Cambridge university press, 2003

  13. [21]

    McMahan, B., Moore, E., Ramage, D., Hampson, S., and y Arcas, B. A. Communication-efficient learning of deep networks from decentralized data. In Artificial intelligence and statistics, pp.\ 1273--1282. PMLR, 2017

  14. [22]

    Murphy, K. P. Machine learning: a probabilistic perspective. MIT press, 2012

  15. [23]

    Pytorch: An imperative style, high-performance deep learning library

    Paszke, A., Gross, S., Massa, F., Lerer, A., Bradbury, J., Chanan, G., Killeen, T., Lin, Z., Gimelshein, N., Antiga, L., et al. Pytorch: An imperative style, high-performance deep learning library. Advances in neural information processing systems, 32, 2019

  16. [24]

    M., and Harchaoui, Z

    Pillutla, K., Kakade, S. M., and Harchaoui, Z. Robust Aggregation for Federated Learning . IEEE Transactions on Signal Processing, 70: 0 1142--1154, 2022. ISSN 1053-587X, 1941-0476. doi:10.1109/TSP.2022.3153135

  17. [25]

    Back to the drawing board: A critical evaluation of poisoning attacks on production federated learning

    Shejwalkar, V., Houmansadr, A., Kairouz, P., and Ramage, D. Back to the drawing board: A critical evaluation of poisoning attacks on production federated learning. In 2022 IEEE Symposium on Security and Privacy (SP), pp.\ 1354--1371, 2022. doi:10.1109/SP46214.2022.9833647

  18. [26]

    L., Jin, L., Liu, N., Li, S., Miao, D., Zhang, X., Ng, W

    Teo, Z. L., Jin, L., Liu, N., Li, S., Miao, D., Zhang, X., Ng, W. Y., Tan, T. F., Lee, D. M., Chua, K. J., et al. Federated machine learning in healthcare: A systematic review on clinical applications and technical architecture. Cell Reports Medicine, 5 0 (2), 2024

  19. [27]

    Privacy preservation in federated learning: An insightful survey from the gdpr perspective

    Truong, N., Sun, K., Wang, S., Guitton, F., and Guo, Y. Privacy preservation in federated learning: An insightful survey from the gdpr perspective. Computers & Security, 110: 0 102402, 2021. ISSN 0167-4048. doi:https://doi.org/10.1016/j.cose.2021.102402. URL https://www.scienc...

  20. [28]

    D., Mayer, R., and Jacobsen, H.-A

    Woisetschläger, H., Erben, A., Marino, B., Wang, S., Lane, N. D., Mayer, R., and Jacobsen, H.-A. Federated learning priorities under the european union artificial intelligence act, 2024. URL https://arxiv.org/abs/2402.05968

  21. [29]

    Fashion-mnist: a novel image dataset for benchmarking machine learning algorithms

    Xiao, H., Rasul, K., and Vollgraf, R. Fashion-mnist: a novel image dataset for benchmarking machine learning algorithms. CoRR, abs/1708.07747, 2017. URL http://arxiv.org/abs/1708.07747

  22. [30]

    Edgefed: Optimized federated learning based on edge computing

    Ye, Y., Li, S., Liu, F., Tang, Y., and Hu, W. Edgefed: Optimized federated learning based on edge computing. IEEE Access, 8: 0 209191--209198, 2020

  23. [31]

    Byzantine-robust distributed learning: Towards optimal statistical rates

    Yin, D., Chen, Y., Kannan, R., and Bartlett, P. Byzantine-robust distributed learning: Towards optimal statistical rates. In International conference on machine learning, pp.\ 5650--5659. Pmlr, 2018

  24. [32]

    A3FL : Adversarially Adaptive Backdoor Attacks to Federated Learning

    Zhang, H., Jia, J., Chen, J., Lin, L., and Wu, D. A3FL : Adversarially Adaptive Backdoor Attacks to Federated Learning . Advances in Neural Information Processing Systems, 36: 0 61213--61233, December 2023 a

  25. [33]

    FLIP : A Provable Defense Framework for Backdoor Mitigation in Federated Learning , February 2023 b

    Zhang, K., Tao, G., Xu, Q., Cheng, S., An, S., Liu, Y., Feng, S., Shen, G., Chen, P.-Y., Ma, S., and Zhang, X. FLIP : A Provable Defense Framework for Backdoor Mitigation in Federated Learning , February 2023 b

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.