Pith. sign in

REVIEW 4 major objections 5 minor 68 references

FedStrategist: A Meta-Learning Framework for Adaptive and Robust Aggregation in Federated Learning

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read FedStrategist argues that the best defense in federated learning is an adaptive one: a contextual bandit chooses among aggregation rules in real time.

desk verdict A plausible adaptive-aggregation idea that its own reported tables contradict: the Krum-champion and lambda-controllability claims do not survive contact with the data. read the letter →

arxiv 2507.14322 v2 pith:FG7U2W3J submitted 2025-07-18 cs.LG cs.CRcs.DC

Md Rafid Haque , Abu Raihan Mostofa Kamal , Md. Azam Hossain This is my paper · ORCID
classification cs.LGcs.CRcs.DC
keywords federatedlearningmodelpoisoningattacksByzantine-robustaggregationcontextualbanditLinUCBadaptivedefenserisktoleranceruleselection
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 argues that no static aggregation rule is best in all federated-learning conditions and that defense selection should itself be adaptive. It introduces FedStrategist, a lightweight contextual bandit that reads cheap diagnostics from the client updates each round and chooses among FedAvg, coordinate-wise median, and Krum. The central claim is that this agent learns policies that match or exceed the best static baselines under standard poisoning, tracks Krum when Krum is favored by the environment, and still keeps a stable, integrity-oriented policy under a stealth attack that blinds one of its diagnostics. The authors further claim that a single risk-tolerance parameter in the reward makes the trade-off between raw accuracy and robust behavior controllable in practice.

What carries the argument

The load-bearing mechanism is a contextual bandit whose reward is $R_t = (\mathrm{Acc}_t - \mathrm{Acc}_{t-1}) - \lambda_{\mathrm{cost}} C_j$, with accuracy measured on a server-held proxy validation set and $C_j$ a normalized heuristic cost for the chosen aggregation rule. LinUCB models each rule's expected reward as a linear function of a three-dimensional state vector—variance of update norms, average pairwise cosine similarity, and norm of the mean update—and selects the rule maximizing the upper confidence bound. This machinery makes 'which defense to run' a per-round decision, and lets $\lambda$ act as a risk dial that the practitioner can turn without retraining the bandit.

What would settle it

Run the stealth-attack experiment to completion, then inspect the final adaptive-agent model for the attacker's injected behavior using trigger patterns or a target-class misclassification probe. If the final model is as compromised as the undefended FedAvg model, or if the agent still selects FedAvg in roughly 94% of rounds as reported in Table 4, the claim that the agent learns a conservative integrity-preserving policy is falsified.

Watch

Extended reading notes

Core claim

The central discovery is that robust aggregation in federated learning is better framed as a real-time control problem than as a choice of a single defense. Using a LinUCB contextual bandit over a three-metric diagnostic state, the paper shows empirically that the adaptive policy outperforms fixed FedAvg, median, and Krum under a standard poisoning attack, tracks Krum when the environment favors Krum, and under a stealth poisoning attack produces a more stable model than the robust baselines while not chasing the highest raw accuracy achieved by a compromised FedAvg. The authors interpret this as evidence that the agent learns a conservative policy that prioritizes model integrity over raw test-set accuracy, and they show that the reward coefficient $\lambda$ shifts the policy between aggressive FedAvg-heavy and cautious median-heavy behavior.

Load-bearing premise

The whole integrity claim rests on the assumption that an increase in validation accuracy from one round to the next means the chosen defense is keeping the model healthy; if a poisoned model can score high on that same accuracy measure, the reward will reward the attacker.

Editorial extensions

If this is right

  • Deploying a single static aggregation rule is a losing strategy in heterogeneous or adversarial settings; operators should expect the best rule to change with the data distribution and threat profile.
  • A single scalar risk tolerance can trade raw accuracy for stability and robustness, so the same framework can serve both aggressive and security-conservative deployments without redesign.
  • The agent can discover and track the best static defense even when that defense is Krum, which suggests the adaptive policy does not bake in a permanent preference for one aggregation family.
  • Lightweight diagnostics are sufficient for the agent to retain a useful policy even when an adversary deliberately neutralizes one of its metrics, lowering the computational cost of deploying adaptive defenses.

Reading between the lines

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

  • The paper leaves implicit that its reward is only as trustworthy as the proxy validation accuracy: the paper's own Table 4 reports the agent selecting FedAvg in 94% of stealth-attack rounds, which suggests that when a poisoned model scores high on that proxy the agent will favor the non-robust rule despite the claimed conservative policy.
  • A natural extension is to replace the accuracy-only reward with an integrity-aware signal, such as a small calibration set probed for backdoor triggers or class-specific corruption, which would let the risk parameter genuinely control robustness rather than mainly control convergence speed.
  • Because the state vector has only three metrics, an adversary who learns to spoof cosine similarity while keeping norm variance low could blind the remaining signals; testing the agent against adversaries that manipulate multiple diagnostics simultaneously would define the real boundary of the framework.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper introduces FedStrategist, a framework that uses a LinUCB contextual bandit to dynamically select among three aggregation rules (FedAvg, coordinate-wise median, Krum) in each federated learning round, based on a three-dimensional diagnostic state vector. The reward is the change in proxy-validation accuracy minus a cost term scaled by a risk parameter λ. The authors claim that the learned policy outperforms all static baselines across diverse scenarios, remains effective under a stealth poisoning attack, and is controllable via λ. The paper includes reproducibility artifacts (code and data on Zenodo).

Significance. The idea of treating robust aggregation as a contextual bandit problem is timely and extends prior work (e.g., SARA) with a lightweight, cost-aware formulation. If the empirical claims were valid, the framework would offer a practical and tunable defense for federated learning. The paper also provides a clear description of the threat model and the reward design. However, the reported experimental tables contradict the main claims: the 'Krum-favorable' experiment shows Krum as the worst static rule, and the λ-sweep shows no policy change for λ ≥ 0.5. Because these contradictions undermine the central contributions, the current results do not support the conclusions.

major comments (4)
  1. [Section 6.4 / Table 3] The text claims 'In this environment, Krum is the clear champion among static rules,' but Table 3 reports Krum's final accuracy as 39.97%, the lowest among FedAvg (43.29%), Median (44.39%), and Adaptive (44.31%). The adaptive agent's performance is nearest to Median, not to Krum. This discrepancy invalidates the claimed validation of the 'generalist strategy' and suggests the experimental results do not support the narrative.
  2. [Section 6.5 / Table 4] The claim that λ acts as an interpretable 'risk dial' is not supported by the data. The reward in Eq. (4.5) includes λ explicitly, so one would expect the policy to shift with λ, but Table 4 shows identical action distributions (94% FedAvg, 4% Median, 2% Krum) for λ = 0.5, 1.0, and 2.0. Only λ = 0.1 differs (92/4/4). This contradicts the statement in Section 7.3 that 'when λ is high (risk-averse), its policy converges to that of the safer, more stable Median.' The agent continues to select FedAvg almost always, even under high λ.
  3. [Section 4.5 and Section 7.2] The reward function R_t = (Acc_t − Acc_{t−1}) − λ·C_j is misaligned with the stated goal of prioritizing model integrity. Section 7.2's 'Paradox of the Losing Winner' explicitly states that the compromised FedAvg can achieve the highest raw accuracy. Consequently, the accuracy-delta term rewards a compromised model. The paper claims the agent 'learned a conservative policy that prioritized model integrity' (Section 7.2), but Table 4 shows the agent chooses FedAvg in 94% of rounds, even at λ = 2.0. Thus the reward does not implement the intended risk management.
  4. [Section 6.3] The claim that the adaptive agent 'outperformed the best static robust aggregator (Median)' is technically true but incomplete. Under the stealth attack (Table 2), the agent's final accuracy (38.82%) is lower than that of FedAvg (39.43%), and the agent's policy is dominated by FedAvg (94% of rounds per Table 4). Therefore the agent is not learning a robust or conservative strategy; it is approximately following the non-robust baseline. This contradicts the interpretation in Section 7.2 that the agent 'identified a persistent attack and prioritized safety.'
minor comments (5)
  1. [Section 6.2, Fig. 4 caption] Both panels are labeled β = 0.5; the top panel should presumably be β = 0.1 given the text in Section 6.2.
  2. [Abstract and Section 6.5] The abstract and Section 6.5 state that controllability is 'proved,' but the paper provides only empirical bar charts and a table; no formal theorem or proof is given.
  3. [Section 4.4] The state vector contains both the variance of update norms and the mean update norm, which are likely correlated; the paper does not discuss whether this redundancy affects LinUCB's linear model.
  4. [Section 5 and Section 6] All experiments use CIFAR-10 and a single CNN; the generality of the conclusions would be strengthened by additional datasets or models.
  5. [Fig. 5 caption] Fig. 5 does not state the heterogeneity parameter β; adding it would improve reproducibility of the stealth-attack experiment.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the lambda-controllability issue is an unsupported empirical claim, not a by-construction reduction.

full rationale

FedStrategist's central claims—that a contextual-bandit selector can beat static aggregators and that its behavior can be tuned by lambda—are empirical claims about a learned policy, not identities derived from the problem setup. The reward R_t = (Acc_t - Acc_{t-1}) - lambda*C_j defines an optimization objective; whether LinUCB actually learns a policy that tracks Krum in the Krum-favorable environment, or changes its action distribution when lambda is varied, is a factual question the experiments were designed to answer. The fact that Table 4 shows the same 94/4/2 distribution at lambda = 0.5, 1.0, and 2.0 undermines the paper's controllability narrative, but that is an internal inconsistency and an unsupported empirical conclusion, not a circular derivation. The only self-citation is reference [64], the authors' code archive, used for reproducibility and not load-bearing. No uniqueness theorem, ansatz, or fitted parameter is imported from prior work. The 'Paradox of the Losing Winner' discussion in Section 7.2 actually acknowledges the reward-signal limitation rather than hiding it. Therefore, under the circularity-specific criteria, no step in the claimed derivation chain reduces to its own input.

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

No invented physical or algorithmic entities are introduced. The free parameters are the risk tolerance lambda, the hand-set defense costs, and the exploration parameter. The axioms are standard assumptions about the reward model and state sufficiency.

free parameters (3)
  • lambda_cost = varied over {0.1, 0.5, 1.0, 2.0}
    User-set risk tolerance in the reward function; central to the claimed controllability result. Not fitted, but the paper's claim that it controls the policy is unsupported by Table 4.
  • C_j (heuristic defense costs) = C_FedAvg=0.1, C_Median=0.4, C_Krum=0.8
    Hand-assigned per-rule costs that directly enter the reward and influence action selection; no grounding or sensitivity analysis provided.
  • alpha (LinUCB exploration parameter) = 1.5
    Exploration-exploitation constant chosen for the bandit; no sensitivity analysis reported.
assumptions (3)
  • domain assumption LinUCB's linear reward model is a valid approximation for the reward of each aggregation rule as a function of the state vector.
    Invoked in Section 4.5; if the true reward surface is strongly nonlinear, the UCB selection will be systematically biased.
  • domain assumption The proxy validation set accuracy delta is a reliable reward signal for model quality and integrity.
    Used in the reward function (Section 4.5). The paper's own 'losing winner' paradox (Section 7.2) shows raw accuracy can favor the compromised model, so this assumption is not guaranteed.
  • domain assumption The 3-dimensional diagnostic state (norm variance, mean cosine similarity, mean norm) contains sufficient information to distinguish attack state from benign heterogeneity.
    Foundation of the instrumentation layer (Section 4.4). If the state is insufficient, no learned policy can be optimal.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FedStrategist: A Meta-Learning Framework for Adaptive and Robust Aggregation in Federated Learning." pith.science (2026). https://pith.science/paper/FG7U2W3J

@misc{pith2026250714322,
  author       = {Pith},
  title        = {Pith review of: FedStrategist: A Meta-Learning Framework for Adaptive and Robust Aggregation in Federated Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FG7U2W3J}},
  note         = {Machine review of arXiv:2507.14322}
}
read the original abstract

Federated Learning (FL) offers a paradigm for privacy-preserving collaborative AI, but its decentralized nature creates significant vulnerabilities to model poisoning attacks. While numerous static defenses exist, their effectiveness is highly context-dependent, often failing against adaptive adversaries or in heterogeneous data environments. This paper introduces FedStrategist, a novel meta-learning framework that reframes robust aggregation as a real-time, cost-aware control problem. We design a lightweight contextual bandit agent that dynamically selects the optimal aggregation rule from an arsenal of defenses based on real-time diagnostic metrics. Through comprehensive experiments, we demonstrate that no single static rule is universally optimal. We show that our adaptive agent successfully learns superior policies across diverse scenarios, including a ``Krum-favorable" environment and against a sophisticated "stealth" adversary designed to neutralize specific diagnostic signals. Critically, we analyze the paradoxical scenario where a non-robust baseline achieves high but compromised accuracy, and demonstrate that our agent learns a conservative policy to prioritize model integrity. Furthermore, we prove the agent's policy is controllable via a single "risk tolerance" parameter, allowing practitioners to explicitly manage the trade-off between performance and security. Our work provides a new, practical, and analyzable approach to creating resilient and intelligent decentralized AI systems.

Figures

Figures reproduced from arXiv: 2507.14322 by the authors.

Figure 1
Figure 1. The Federated Learning Aggregation Trilemma. Static aggregation rules force a trade-off between performance (FedAvg), robustness to attacks (Krum), and robustness to data heterogeneity (Median). The FedStrategist framework is designed to dynamically navigate this trade-off space by selecting the optimal rule for the current network conditions. 3 Related Work Our research is situated at the confluence of three major … view at source ↗
Figure 2
Figure 2. The FedStrategist Framework Architecture. In each round (t), (1) the server collects updates (∆wi) from clients. (2) The Instrumentation Layer computes a state vector (St). (3) The Meta-Learning Agent (Contextual Bandit) ingests St and (4) chooses an aggregation rule ’j’ from the Defense Arsenal. (5) The chosen rule is applied to the updates to compute the new global model Wt. (6) A reward Rt is calculated based on … view at source ↗
Figure 3
Figure 3. Conceptual Visualization of the Learned Policy. This diagram illustrates the idealized decision regions. The agent is expected to learn to select the aggressive ’FedAvg’ in safe conditions (low variance, high similarity), the geometric ’Krum’ against loud attacks (high variance), and the statistical ’Median’ when faced with high data heterogeneity (low similarity). September 19, 2025 10/23 [PITH_FULL_IMAGE:figures/… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Model accuracy under a standard poisoning attack. Top: High heterogeneity (β = 0.5). Bottom: Moderate heterogeneity (β = 0.5). The FedStrategist (’adaptive’) agent learns a policy that surpasses all static baselines. September 19, 2025 11/23 [PITH_FULL_IMAGE:figures/f…
Figure 5
Figure 5. Figure 5: Model accuracy under a stealth poisoning attack (β = 0.5). The attack’s regularizing effect creates a complex trade-off between the compromised ’FedAvg’ and the safer robust methods. The results, shown in [PITH_FULL_IMAGE:figures/full_fig_p012_5.png]
Figure 6
Figure 6. Figure 6: Model accuracy in a Krum-favorable environment (β = 10.0, standard attack). The adaptive agent correctly identifies and tracks the performance of ’Krum’ as the optimal static strategy. true generalist, capable of learning the optimal defensive posture based on the spec…
Figure 7
Figure 7. Figure 7: Impact of the λcost parameter on agent performance and strategy under the stealth attack. (Left) Accuracy curves for different λ values, demonstrating a clear shift from risk-seeking (tracking ’FedAvg’) to more conservative behavior. (Right) The agent’s chosen rule per…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

68 extracted references · 64 canonical work pages

  1. [1]

    Communication-Efficient Learning of Deep Networks from Decentralized Data

    McMahan HB, Moore E, Ramage D, Hampson S, Ag¨ uera y Arcas B. Communication-Efficient Learning of Deep Networks from Decentralized Data. In: Gretton A, Robert CC, editors. Proceedings of the 20th International Conference on Artificial Intelligence and Statistics. vol. 54 of Proceedings of Machine Learning Research. Fort Lauderdale, FL, USA: PMLR; 2017. p....

    work page 2017

  2. [2]

    Federated Learning: Challenges, Methods, and Future Directions

    Li T, Sahu AK, Talwalkar A, Smith V. Federated Learning: Challenges, Methods, and Future Directions. IEEE Signal Processing Magazine. 2020;37(3):50–60

    work page 2020

  3. [3]

    Challenges, Applications and Design Aspects of Federated Learning: A Survey

    Rahman KMJ, Ahmed F, Akhter N, Hasan M, Amin R, Aziz KE, et al. Challenges, Applications and Design Aspects of Federated Learning: A Survey. IEEE Access. 2021;9:124682–124700. doi:10.1109/ACCESS.2021.3111118

    work page arXiv 2021

  4. [4]

    Self-Sovereign Identity Management for Hierarchical Federated Learning in Vehicular Networks

    Zeydan E, Mangues J, Arslan SS, Turk Y. Self-Sovereign Identity Management for Hierarchical Federated Learning in Vehicular Networks. In: 2023 International Conference on High Performance Switching and Routing (HPSR). Rio de Janeiro, Brazil: IEEE; 2023

    work page 2023

  5. [5]

    Local Model Poisoning Attacks to Byzantine-Robust Federated Learning

    Fang M, Cao X, Jia J, Gong N. Local Model Poisoning Attacks to Byzantine-Robust Federated Learning. In: USENIX Security Symposium. USENIX Security 20. Santa Clara, CA, USA: USENIX Association; 2020. p. 1605–1622

    work page 2020

  6. [6]

    Manipulating the Byzantine: Optimizing Model Poisoning Attacks and Defenses for Federated Learning

    Shejwalkar V, Houmansadr A. Manipulating the Byzantine: Optimizing Model Poisoning Attacks and Defenses for Federated Learning. In: Proceedings of the Network and Distributed System Security Symposium (NDSS). The Internet Society; 2021

    work page 2021

  7. [7]

    How To Backdoor Federated Learning

    Bagdasaryan E, Veit A, Hua Y, Estrin D, Shmatikov V. How To Backdoor Federated Learning. In: Chiappa S, Calandra R, editors. Proceedings of the 23rd International Conference on Artificial Intelligence and Statistics. vol. 108 of Proceedings of Machine Learning Research. Palermo, Italy: PMLR; 2020. p. 2938–2948

    work page 2020

  8. [8]

    Free-riders in Federated Learning: Attacks and Defenses

    Lin J, Du M, Liu J. Free-riders in Federated Learning: Attacks and Defenses

    work page

Show all 68 references
  1. [9]

    Blockchain for Federated Learning Toward Secure Distributed Machine Learning Systems: A Systemic Survey

    Li D, Han D, Weng TH, Zheng Z, Li H, Liu H, et al. Blockchain for Federated Learning Toward Secure Distributed Machine Learning Systems: A Systemic Survey. Soft Computing. 2022;26(9):4423–4440

    2022

  2. [10]

    A Blockchain-based Trust System for Decentralised Applications: When trustless needs trust

    Truong N, Lee GM, Sun K, Guitton F, Guo Y. A Blockchain-based Trust System for Decentralised Applications: When trustless needs trust. arXiv preprint arXiv:210110920. 2021

    2021

  3. [11]

    Blockchain-Enabled Federated Learning: A Reference Architecture Design, Implementation, and Verification

    Goh E, Kim DY, Lee K, Oh S, Chae JE, Kim DY. Blockchain-Enabled Federated Learning: A Reference Architecture Design, Implementation, and Verification. IEEE Access. 2023;11:53934–53949

    2023

  4. [12]

    Biscotti: A Blockchain System for Private and Secure Federated Learning

    Shayan M, Fung C, Yoon CJM, Beschastnikh I. Biscotti: A Blockchain System for Private and Secure Federated Learning. IEEE Transactions on Parallel and Distributed Systems. 2021;32(7):1513–1525

    2021

  5. [13]

    Proof-of-Reputation: An Alternative Consensus Mechanism for Blockchain Systems

    Aluko O, Kolonin A. Proof-of-Reputation: An Alternative Consensus Mechanism for Blockchain Systems. International Journal of Network Security & Its Applications. 2021;13(4):23–40. doi:10.5121/ijnsa.2021.13403. September 19, 2025 19/23

    2021

  6. [14]

    MeritRank: Sybil Tolerant Reputation for Merit-based Tokenomics

    Nasrulin B, Ishmaev G, Pouwelse J. MeritRank: Sybil Tolerant Reputation for Merit-based Tokenomics. In: 2022 IEEE International Conference on Blockchain and Cryptocurrency (ICBC). Shanghai, China: IEEE; 2022. p. 1–9

    2022

  7. [15]

    Efficiency in Digital Economies – A Primer on Tokenomics; 2020

    Lamberty R, Poddey A, Galindo D, de Waard D, Koelbel T, Kirste D. Efficiency in Digital Economies – A Primer on Tokenomics; 2020. arXiv preprint arXiv:2008.02538

    2020 arXiv

  8. [16]

    Engineering Token Economy with System Modeling

    Zhang Z. Engineering Token Economy with System Modeling. arXiv preprint arXiv:190700899. 2019

    2019

  9. [17]

    Computational Attestations of Polynomial Integrity Towards Verifiable Machine Learning; 2024

    Ray D, Jazmi CE. Computational Attestations of Polynomial Integrity Towards Verifiable Machine Learning; 2024. Cryptology ePrint Archive, Paper 2024/639. Available from: https://eprint.iacr.org/2024/639

    2024

  10. [18]

    ZHE: Efficient Zero-Knowledge Proofs for HE Evaluations; 2025

    Zhou Z, Li Y, Wang Y, Yang Z, Zhu Y, Chen Y, et al.. ZHE: Efficient Zero-Knowledge Proofs for HE Evaluations; 2025. Cryptology ePrint Archive, Paper 2025/770. Available from: https://eprint.iacr.org/2025/770

    2025

  11. [19]

    Machine Learning with Adversaries: Byzantine Tolerant Gradient Descent

    Blanchard P, Mhamdi EME, Guerraoui R, Stainer J. Machine Learning with Adversaries: Byzantine Tolerant Gradient Descent. In: Guyon I, Luxburg UV, Bengio S, Wallach H, Fergus R, Vishwanathan S, et al., editors. Advances in Neural Information Processing Systems 30. NIPS 2017. Lo...

    2017

  12. [20]

    Automatic Adversarial Adaption for Stealthy Poisoning Attacks in Federated Learning

    Krauß T, K¨ onig J, Dmitrienko A, Kanzow C. Automatic Adversarial Adaption for Stealthy Poisoning Attacks in Federated Learning. In: Proceedings of the Network and Distributed System Security Symposium (NDSS). San Diego, California, USA: Internet Society; 2024

    2024

  13. [21]

    Developing Hessian–Free Second–Order Adversarial Examples for Adversarial Training

    Qian Y, Zhang L, Wang Y, Ji B, Yao T, Wang B. Developing Hessian–Free Second–Order Adversarial Examples for Adversarial Training. International Journal of Applied Mathematics and Computer Science. 2024;34(3):425–438. doi:10.61822/amcs-2024-0030

    2024 doi

  14. [22]

    Federated Learning: Overview, strategies, applications, tools and future directions

    Yurdem B, Kuzlu M, Gullu MK, Catak FO, Tabassum M. Federated Learning: Overview, strategies, applications, tools and future directions. Heliyon. 2024;10(19):e38137

    2024

  15. [23]

    Federated Learning: A Distributed Shared Machine Learning Method

    Hu K, Li Y, Xia M, Wu J, Lu M, Zhang S, et al. Federated Learning: A Distributed Shared Machine Learning Method. Complexity. 2021

    2021

  16. [24]

    Federated Learning: Strategies for Improving Communication Efficiency

    Koneˇ cn´ y J, McMahan HB, Yu FX, Suresh AT, Bacon D, Richt´ arik P. Federated Learning: Strategies for Improving Communication Efficiency. arXiv preprint arXiv:161005492. 2016

    2016

  17. [25]

    Federated Averaging: The Backbone of Federated Learning; 2024

    Team R. Federated Averaging: The Backbone of Federated Learning; 2024. Available from: https://www.rtinsights.com/ federated-averaging-the-backbone-of-federated-learning/

    2024

  18. [26]

    Limitations and Future Aspects of Communication Costs in Federated Learning: A Survey

    Authors V. Limitations and Future Aspects of Communication Costs in Federated Learning: A Survey. Sensors. 2023;23(17):7358

    2023

  19. [27]

    Exploring the Practicality of Federated Learning: A Survey Towards the Communication Perspective

    Le K, Luong-Ha N, Nguyen-Duc M, Le-Phuoc D, Do C, Wong KS. Exploring the Practicality of Federated Learning: A Survey Towards the Communication Perspective. arXiv preprint arXiv:240520431. 2024;. September 19, 2025 20/23

    2024

  20. [28]

    Communication-Efficient Federated Learning for Resource-Constrained Edge Devices

    Lan G, Liu XY, Zhang Y, Wang X. Communication-Efficient Federated Learning for Resource-Constrained Edge Devices. IEEE Transactions on Machine Learning in Communications and Networking. 2023

    2023

  21. [29]

    Federated Learning with Non-IID Data

    Zhao Y, Li M, Lai L, Suda N, Civin D, Chandra V. Federated Learning with Non-IID Data. arXiv preprint arXiv:180600582. 2018

    2018

  22. [30]

    A Survey on Heterogeneous Federated Learning

    Gao D, Yao X, Yang Q. A Survey on Heterogeneous Federated Learning. arXiv preprint arXiv:221004505. 2022

    2022

  23. [31]

    Heterogeneous Federated Learning: State-of-the-art and Research Challenges

    Fang X, Ye M, Du B, Yuen PC, Tao D. Heterogeneous Federated Learning: State-of-the-art and Research Challenges. ACM Computing Surveys. 2023

    2023

  24. [32]

    The Limitations of Federated Learning in Sybil Settings

    Fung C, Yoon CJM, Beschastnikh I. The Limitations of Federated Learning in Sybil Settings. In: 23rd International Symposium on Research in Attacks, Intrusions and Defenses (RAID 2020). San Sebastian: USENIX Association; 2020. p. 301–316. Available from: https://www.usenix.org/...

    2020

  25. [33]

    DMPA: Model Poisoning Attacks on Decentralized Federated Learning for Model Differences; 2025

    Feng C, Li Y, Gao Y, Celdr´ an AH, von der Assen J, Bovet G, et al.. DMPA: Model Poisoning Attacks on Decentralized Federated Learning for Model Differences; 2025. arXiv preprint arXiv:2502.04771

    2025 arXiv

  26. [34]

    Securing Federated Learning Against Overwhelming Collusive Attackers

    Ranjan P, Gupta A, Cor` o F, Das SK. Securing Federated Learning Against Overwhelming Collusive Attackers. In: IEEE Global Communications Conference (GLOBECOM). Rio de Janeiro, Brazil: IEEE; 2022

    2022

  27. [35]

    Deep Leakage from Gradients

    Zhu L, Liu Z, Han S. Deep Leakage from Gradients. In: Advances in Neural Information Processing Systems. vol. 32 of NeurIPS ’19. Curran Associates, Inc

  28. [36]

    A Survey of Distributed Consensus Protocols for Blockchain Networks

    Xiao Y, Zhang N, Lou W, Hou YT. A Survey of Distributed Consensus Protocols for Blockchain Networks. IEEE Communications Surveys & Tutorials. 2020;22(2):1432–1465. doi:10.1109/COMST.2020.2969706

    2020

  29. [37]

    Consensus Algorithms of Distributed Ledger Technology – A Comprehensive Analysis; 2023

    Alkhodair AJ, Mohanty SP, Kougianos E. Consensus Algorithms of Distributed Ledger Technology – A Comprehensive Analysis; 2023. arXiv preprint arXiv:2309.13498

    2023 arXiv

  30. [38]

    The Byzantine Generals Problem

    Lamport L, Shostak R, Pease M. The Byzantine Generals Problem. ACM Transactions on Programming Languages and Systems. 1982;4(3):382–401

    1982

  31. [39]

    Practical Byzantine Fault Tolerance

    Castro M, Liskov B. Practical Byzantine Fault Tolerance. In: Proceedings of the Third Symposium on Operating Systems Design and Implementation. OSDI ’99. New Orleans, LA, USA: USENIX Association; 1999. p. 173–186

    1999

  32. [40]

    HotStuff: BFT Consensus with Linearity and Responsiveness

    Yin M, Malkhi D, Reiter M, Golan-Gueta G, Abraham I. HotStuff: BFT Consensus with Linearity and Responsiveness. In: Proceedings of the 2019 ACM SIGACT–SIGOPS Symposium on Principles of Distributed Computing. PODC ’19. Toronto, ON, Canada: Association for Computing Machinery; 2...

    2019

  33. [41]

    Security Analysis Methods on Ethereum Smart Contract Vulnerabilities: A Survey; 2019

    Praitheeshan P, Pan L, Yu J, Liu JK, Doss R. Security Analysis Methods on Ethereum Smart Contract Vulnerabilities: A Survey; 2019. arXiv preprint arXiv:1908.08605. September 19, 2025 21/23

    2019 arXiv

  34. [42]

    Beyond the Tragedy of the Commons: Building A Reputation System for Generative Multi-agent Systems

    Ren S, Fu W, Zou X, Shen C, Cai Y, Chu C, et al.. Beyond the Tragedy of the Commons: Building A Reputation System for Generative Multi-agent Systems

  35. [43]

    DID-eFed: Facilitating Federated Learning as a Service with Decentralized Identities

    Geng J, Kanwal N, Jaatun MG, Rong C. DID-eFed: Facilitating Federated Learning as a Service with Decentralized Identities. In: Proceedings of the 25th International Conference on Evaluation and Assessment in Software Engineering. EASE ’21. Trondheim, Norway: ACM; 2021. p. 123–130

    2021

  36. [44]

    Cryptoeconomics and Tokenomics as Economics: A Survey with Opinions

    Ito K. Cryptoeconomics and Tokenomics as Economics: A Survey with Opinions

  37. [45]

    In: Proceedings of the 11th International Conference on Autonomous Agents and Multiagent Systems

    Token Economy for Online Exchange Systems. In: Proceedings of the 11th International Conference on Autonomous Agents and Multiagent Systems. AAMAS ’12. International Foundation for Autonomous Agents and Multiagent Systems; 2012

    2012

  38. [46]

    Using Game Theory to Design Resilient Token Economies; 2024

    fi Team T. Using Game Theory to Design Resilient Token Economies; 2024. TDE.fi Blog. Available from: https://tde.fi/founder-resource/blogs/ tokenomics/using-game-theory-to-design-resilient-token-economies/

    2024

  39. [47]

    Applying Game Theory in Token Design; 2024

    Team N. Applying Game Theory in Token Design; 2024. Nextrope Blog. Available from: https://nextrope.com/applying-game-theory-in-token-design/

    2024

  40. [48]

    Tokenomics and Game Theory: Understanding the Economic Incentives of Blockchain Networks; 2024

    Contributor L. Tokenomics and Game Theory: Understanding the Economic Incentives of Blockchain Networks; 2024. LinkedIn Pulse. Available from: https://www.linkedin.com/pulse/ tokenomics-game-theory-understanding-economic-incentives-wbble

    2024

  41. [49]

    Game Theory and Blockchain; 2025

    Team C. Game Theory and Blockchain; 2025. Coinsquare Learn. Available from: https://www.coinsquare.com/en-ca/learn/game-theory-and-blockchain

    2025

  42. [50]

    Incentive Compatibility in Consensus Protocols and DAOs: A Game-Theoretic Approach

    Murano A, Bruno B, Vespri V. Incentive Compatibility in Consensus Protocols and DAOs: A Game-Theoretic Approach. In: CEUR Workshop Proceedings. vol

  43. [51]

    Game Theory-Based Incentive Design in Blockchain Networks

    Contributors E. Game Theory-Based Incentive Design in Blockchain Networks. Encyclopedia. 2024

    2024

  44. [52]

    A Survey on Homomorphic Encryption Schemes: Theory and Implementation

    Acar A, Aksu H, Uluagac AS, Conti M. A Survey on Homomorphic Encryption Schemes: Theory and Implementation. arXiv preprint arXiv:170403578. 2017

    2017

  45. [53]

    PBFL: A Privacy-Preserving Blockchain-Based Federated Learning Framework With Homomorphic Encryption and Single Masking

    Han B, Li B, Jurdak R, Zhang P. PBFL: A Privacy-Preserving Blockchain-Based Federated Learning Framework With Homomorphic Encryption and Single Masking. IEEE Internet of Things Journal. 2025

    2025

  46. [54]

    A Review of Homomorphic Encryption for Privacy-Preserving Computations

    Yang W, Wang S, Cui N, Tan J, et al. A Review of Homomorphic Encryption for Privacy-Preserving Computations. PMC. 2023;10:e98691

    2023

  47. [55]

    Advancing Blockchain-based Federated Learning through Verifiable Off-chain Computations

    Heiss J, Grunewald E, Haimerl N, Schulte S, Tai S. Advancing Blockchain-based Federated Learning through Verifiable Off-chain Computations. In: International Congress on Blockchain and Applications. vol. 497 of Lecture Notes in Networks and Systems. Cham: Springer; 2022. p. 170–179

    2022

  48. [56]

    Mathematical Proposal for Securing Split Learning Using Homomorphic Encryption and Zero-Knowledge Proofs

    Kokaj A, Mollakuqe E. Mathematical Proposal for Securing Split Learning Using Homomorphic Encryption and Zero-Knowledge Proofs. Applied Sciences. 2025;15(6):2913. doi:10.3390/app15062913. September 19, 2025 22/23

    2025 doi

  49. [57]

    Byzantine-Robust Distributed Learning: Towards Optimal Statistical Rates

    Yin D, Chen Y, Ramchandran K, Bartlett P. Byzantine-Robust Distributed Learning: Towards Optimal Statistical Rates. In: Dy J, Krause A, editors. Proceedings of the 35th International Conference on Machine Learning. vol. 80 of Proceedings of Machine Learning Research. Stockholm...

    2018

  50. [58]

    Robust Federated Learning: Maximum Correntropy Aggregation Against Byzantine Attacks

    Luan Z, Li W, Liu Mq, Chen B. Robust Federated Learning: Maximum Correntropy Aggregation Against Byzantine Attacks. IEEE Transactions on Neural Networks and Learning Systems. 2024

    2024

  51. [59]

    An Experimental Study of Byzantine-Robust Aggregation Schemes in Federated Learning

    Li S, Ngai ECH, Voigt T. An Experimental Study of Byzantine-Robust Aggregation Schemes in Federated Learning. IEEE Transactions on Big Data. 2024;10(6):975–988

    2024

  52. [60]

    Adapting Aggregation Rule for Robust Federated Learning under Dynamic Attacks

    Hu C, Zhang M, Li N, Li J, Yang Z, Ul Hassan M, et al. Adapting Aggregation Rule for Robust Federated Learning under Dynamic Attacks. In: 2025 IEEE/ACM 20th Symposium on Software Engineering for Adaptive and Self-Managing Systems (SEAMS); 2025. p. 171–177

    2025

  53. [61]

    MAB-RFL: A multi-armed bandit based robust federated learning framework

    Wan W, Hu S, Li M, Lu J, Zhang L, Zhang LY, et al. MAB-RFL: A multi-armed bandit based robust federated learning framework. Journal of Parallel and Distributed Computing. 2021;156:14–24

    2021

  54. [62]

    UCB-CS: A New Bandit-Based Approach for Communication-Efficient Client Selection in Federated Learning

    Cho YY, Lee GY, Poor HV. UCB-CS: A New Bandit-Based Approach for Communication-Efficient Client Selection in Federated Learning. In: 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). Rhodes Island, Greece: IEEE; 2023. p. 1–5

    2023

  55. [63]

    FedCostAware: Enabling Cost-Aware Federated Learning on the Cloud; 2025

    Sinha A, Li Z, Liu T, Kindratenko V, Kim K, Madduri R. FedCostAware: Enabling Cost-Aware Federated Learning on the Cloud; 2025. Available from: https://arxiv.org/abs/2505.21727

    2025 arXiv

  56. [64]

    Fedstrategist; 2025

    Haque MR, Kamal ARM, Hossain MA. Fedstrategist; 2025. Available from: https://doi.org/10.5281/zenodo.16068113. September 19, 2025 23/23

    2025 doi

  57. [2019]

    arXiv preprint arXiv:1911.12560

    1911 arXiv

  58. [2024]

    arXiv preprint arXiv:2407.15715

    arXiv

  59. [2025]

    arXiv preprint arXiv:2501.05029

    arXiv

  60. [3791]

    CEUR-WS.org; 2025. p. 1–10

    2025

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.