Byzantine-Resilient Secure Federated Learning
Open accessThe DOI matches its Crossref record; 47 of the 48 references it cites were found in a registry or an index.
Abstract
Secure federated learning is a privacy-preserving framework to improve machine learning models by training over large volumes of data collected by mobile users. This is achieved through an iterative process where, at each iteration, users update a global model using their local datasets. Each user then masks its local update via random keys, and the masked models are aggregated at a central server to compute the global model for the next iteration. As the local updates are protected by random masks, the server cannot observe their true values. This presents a major challenge for the resilience of the model against adversarial (Byzantine) users, who can manipulate the global model by modifying their local updates or datasets. Towards addressing this challenge, this paper presents the first single-server Byzantine-resilient secure aggregation framework (BREA) for secure federated learning. BREA is based on an integrated stochastic quantization, verifiable outlier detection, and secure model aggregation approach to guarantee Byzantine-resilience, privacy, and convergence simultaneously. We provide theoretical convergence and privacy guarantees and characterize the fundamental trade-offs in terms of the network size, user dropouts, and privacy protection. Our experiments demonstrate convergence in the presence of Byzantine users, and comparable accuracy to conventional federated learning benchmarks.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“Bayesian neural networks [419] have been proposed and shown to be far more robust to over-fitting, and can easily learn from small datasets.”
From Advances and Open Problems in Federated Learning (2021) · cites this paper as [419]
In the citing paper
“To simultaneously preserve privacy for individual users while tolerating Byzantine adversaries, a Byzantine-resilient secure aggregation framework was developed in [171] to detect adversarial models without the knowledge of individual local models, as they are masked for privacy guarantees.”
From Edge artificial intelligence for 6g: vision, enabling technologies, and applications (2022) · cites this paper as [171]
Citations
Cites
Adam: A Method for Stochastic Optimization
Kingma, D. P., & Ba, J. (2014) · arXiv · doi:10.48550/arxiv.1412.6980 · Cited by 98,946
Learning Multiple Layers of Features from Tiny Images
Krizhevsky, A. (2024) · doi:10.57702/zp44cu3g · Cited by 23,395
New directions in cryptography
Diffie, W., & Hellman, M. (1976) · IEEE Transactions on Information Theory · doi:10.1109/tit.1976.1055638 · Cited by 14,713
Communication-Efficient Learning of Deep Networks from Decentralized Data
McMahan, H. B., Moore, E., Ramage, D., Hampson, S., & Arcas, B. A. Y. (2016) · arXiv · doi:10.48550/arxiv.1602.05629 · Cited by 14,215
Shamir, A. (1979) · Communications of the ACM · doi:10.1145/359168.359176 · Cited by 13,756 · cited 5×
“Our framework leverages Feldman’s verifiable secret sharing protocol from [17], which combines Shamir’s secret sharing [18] with homomorphic encryption.”
BackgroundMethodInfluentialAdvances and Open Problems in Federated Learning
Kairouz, P., & McMahan, H. B. (2021) · Foundations and Trends in Machine Learning · doi:10.1561/2200000083 · Cited by 7,772
Federated learning: Challenges, methods, and future directions
Li, T., Sahu, A. K., Talwalkar, A., & Smith, V. (2020) · IEEE Signal Processing Magazine · doi:10.1109/msp.2020.2975749 · Cited by 6,210
Practical Secure Aggregation for Privacy-Preserving Machine Learning
Bonawitz, K., Ivanov, V., Kreuter, B., Marcedone, A., McMahan, H. B., Patel, S., Ramage, D., Segal, A., & Seth, K. (2017) · Proceedings of the 2017 ACM SIGSAC Conference on Computer and Communications Security · doi:10.1145/3133956.3133982 · Cited by 3,956
Federated Learning: Strategies for Improving Communication Efficiency
Konečný, J., McMahan, H. B., Yu, F. X., Richtárik, P., Suresh, A. T., & Bacon, D. (2016) · arXiv · doi:10.48550/arxiv.1610.05492 · Cited by 3,766
Differential Privacy: A Survey of Results
Dwork, C. (2008) · Lecture Notes in Computer Science · doi:10.1007/978-3-540-79228-4_1 · Cited by 3,542 · cited 1×
“Although our focus is not on differential-privacy, our approach may in principle be combined with differential privacy techniques [27], which is an interesting future direction.”
Background
Research works printed but not linked: 3
- bottou, 1998, Online learning and stochastic approximations, Online Learning in Neural Networks, 17, 142
- mcmahan, 2017, Communication-efficient learning of deep networks from decentralized data, Proc 20th Int Conf Artif Intell Statist, 54, 1273
- lecun, 2010, MNIST Handwritten Digit Database
Cited by
Items 301–316 of 316
Information-Agnostic Model Poisoning Attacks Against Byzantine-Robust Federated Learning
Zhang, Y., Chen, Y., Tan, X., Shen, D., Wang, M. Y., & Wang, B. (2026) · Lecture notes in computer science · doi:10.1007/978-981-95-4149-2_20
Wang, Y., Lu, Y., & Ni, Q. (2026) · IEEE Internet of Things Journal · doi:10.1109/jiot.2026.3705543
“However, mainstream FL approaches [7], [8] face fundamental limitations in streaming scenarios.”
Edge Computing with Federated Learning: Privacy and Performance Trade-offs
Dixit, M., & Kumar, R. (2025) · Asian Conference on Communication and Networks (ASIANComNet) · doi:10.1109/asiancomnet68615.2025.11579552
“A part of the research concentrated on the impact of differences in devices on FL performance and suggested an adaptive learning rate to avoid the drop in performance [11].”
Hazem, M., & Zamzam, M. (2026) · International Conference on Machine Intelligence and Smart Innovation (ICMISI) · doi:10.1109/icmisi69868.2026.11584141
“Additionally, FL is vulnerable to Free Rider Attacks in which the attackers exploit the aggregation outcome without sending their authenticated labels [3].”
FeedSign: Robust and Communication-Efficient Federated Fine-tuning of Large Models for Edge AI
Cai, Z., Chen, H., Zhu, G., Shi, Q., & Huang, K. (2026) · International Conference on Communications · doi:10.1109/icc59461.2026.11587130
“Various aggregation methods have been proposed to improve FO-based FL resilience against such attacks [17].”
Stress Monitoring with Machine Learning: Detecting and Managing Mental Health
Jiang, S., Li, Y., Firouzi, F., Chakrabarty, K., & Elbogen, E. B. (2025) · Springer eBooks · doi:10.1007/978-3-032-09253-3_4
Learning-Aware Anomaly Detection in Secure Federated Learning
V., T. N. A., & Singh, M. P. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-35586-7_3
Halenka, V., Amini, M., Andersen, P.-A., Granmo, O.-C., Jiao, L., & Kantarci, B. (2026) · IEEE Transactions on Network Science and Engineering · doi:10.1109/tnse.2026.3727545
FPMAN: A Framework for Stable and Privacy-Preserving Federated Learning under Non-IID Data
Esha, I. T., Tasnim, T., & Das, D. (2026) · International Conference on Power, Electronics, Communications, Computing, and Intelligent Infrastructure (PECCII) · doi:10.1109/peccii70991.2026.11662022
“Secure aggregation methods, such as secret sharing, hide individual client updates during aggregation [11–13], but most approaches treat all network layers equally, even though earlier layers learn general features and deeper layers learn sensitive patterns [14–16].”
Tuama, A. Y., Hasan, T. M., & Shakir, A. C. (2026) · Journal of Information Security and Applications · doi:10.1016/j.jisa.2026.104608
Hussain, T., Azam, F., & Anwar, M. W. (2026) · International Conference on Cloud Computing and Big Data Analytics (ICCCBDA) · doi:10.1109/icccbda69573.2026.11676430
“BREA [10] provides Byzantine resilience with privacy via coded computing, but assumes an honest server majority.”
Rahmati, M., & Rahmati, N. (2026) · Journal of Electrical Systems and Information Technology · doi:10.1186/s43067-026-00401-6
“Updates with magnitudes exceeding a threshold 𝜏𝜏 𝑚𝑚 = 𝜇𝜇 𝑚𝑚 + 𝛽𝛽𝛽𝛽 𝑚𝑚 are flagged as suspicious, where 𝑚𝑚 and 𝑚𝑚 denote the mean and standard deviation of magnitudes across all submitted updates in round 𝑡𝑡 , and is a sensitivity parameter typically set between 2 and 3 [71].”
Xia, Y., Jahani-Nezhad, T., & Bitar, R. (2026) · IEEE Journal on Selected Areas in Communications · doi:10.1109/jsac.2026.3744992 · Cited by 1
“For instance, privacy and robustness are jointly considered in [12], [13], [21], [30], [31], [32], [33], [34], and [35]; communication efficiency and privacy are studied in [16], [36], [37], and [38]; and communication efficiency and robustness are explored in [39], [40], [41], and [42].”
VGPD-FL: Group-level poisoning detection without trusted datasets for secure aggregation
Li, J., Li, Y., Zhou, T., Ke, W., Wang, S., & Yang, X. (2026) · Journal of Information Security and Applications · doi:10.1016/j.jisa.2026.104673
Lin, J., Zhang, L., Tu, J., Ma, W., & Guo, R. (2026) · Journal of Information Security and Applications · doi:10.1016/j.jisa.2026.104696
Metadata
- Authors
- So, J., Guler, B., & Avestimehr, A. S.
- Affiliations
- University of Southern California (US)
- Venue
- IEEE Journal on Selected Areas in Communications
- Published
- 2021-07
- Volume
- 39
- Issue
- 7
- Pages
- 2168-2181
- ISSN
- 0733-8716 (ISSN-L)
- 1558-0008
- OpenAlex
- W3107632872
- Cited by
- 313
- References
- 48
- Citations by year
- 2026: 32 · 2025: 81 · 2024: 67 · 2023: 55 · 2022: 43 · 2021: 27
- FWCI
- 13.06
- Type
- journal article
- Language
- en
- Open access
- green
- arxiv.org
- Licence
- Locations
- IEEE Journal on Selected Areas in Communications (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Funders
- Defense Advanced Research Projects Agency (HR001117C0053)
- Army Research Office (W911NF1810400)
- NSF (CCF-1703575, CCF-1763673)
- Office of Naval Research (N00014-16-1-2189)
- research gifts from Intel and Facebook
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Privacy-Preserving Technologies in Data
- Physical Sciences › Computer Science › Computer Networks and Communications › Cryptography and Data Security
- Physical Sciences › Computer Science › Information Systems › Cloud Data Security Solutions
- Keywords
- secure federated learning, Byzantine resilience, secure aggregation, stochastic quantization, privacy preservation, convergence guarantees, user dropout
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.