{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,15]],"date-time":"2026-06-15T09:52:10Z","timestamp":1781517130635,"version":"3.54.1"},"reference-count":42,"publisher":"Institute for Operations Research and the Management Sciences (INFORMS)","issue":"3","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["INFORMS Journal on Computing"],"published-print":{"date-parts":[[2026,5]]},"abstract":"<jats:p>Federated learning is a novel framework for distributed learning, which aims to break isolated data islands, as well as protect data privacy. To further prevent privacy leakage by specially crafted attacks, differential privacy is often integrated. Although differential privacy effectively secures sensitive information, it can reduce the statistical efficiency of the resulting estimators. This leads to a trade-off relationship between statistical efficiency and privacy protection. To theoretically understand this relationship, we start with the classic linear regression model and a noise-adding federated gradient descent algorithm. Its numerical convergence properties and asymptotic properties are rigorously studied. This results in fruitful insights into the trade-off relationship between statistical efficiency and privacy protection. Guided by these theoretical understandings, we further develop a Polyak-Ruppert-type averaged estimator, which can achieve good statistical efficiency with guaranteed privacy protection. Extensive simulation studies are presented to corroborate our theoretical results. Finally, we illustrate the application of our proposed method on an enterprise community data set.<\/jats:p>\n                  <jats:p>History: Accepted by Ram Ramesh, Area Editor for Data Science &amp; Machine Learning.<\/jats:p>\n                  <jats:p>Funding: Financial support from the National Natural Science Foundation of China [Grants 12401386, 72371241, 72495123, and 12271012], the Ministry of Education Project of Key Research Institute of Humanities and Social Sciences [Grant 22JJD910001], the Postdoctoral Fellowship Program of China Postdoctoral Science Foundation [Grant GZB20230070], and the Beijing Municipal Social Science Foundation [Grant 24GLC033] is gratefully acknowledged.<\/jats:p>\n                  <jats:p>Supplemental Material: The software that supports the findings of this study is available within the paper and its Supplemental Information ( https:\/\/pubsonline.informs.org\/doi\/suppl\/10.1287\/ijoc.2024.0554 ) as well as from the IJOC GitHub software repository ( https:\/\/github.com\/INFORMSJoC\/2024.0554 ). The complete IJOC Software and Data Repository is available at https:\/\/informsjoc.github.io\/ .<\/jats:p>","DOI":"10.1287\/ijoc.2024.0554","type":"journal-article","created":{"date-parts":[[2025,7,15]],"date-time":"2025-07-15T16:42:21Z","timestamp":1752597741000},"page":"943-957","source":"Crossref","is-referenced-by-count":2,"title":["Privacy Protection and Statistical Efficiency Trade-Off for Federated Learning"],"prefix":"10.1287","volume":"38","author":[{"ORCID":"https:\/\/orcid.org\/0009-0007-8128-6360","authenticated-orcid":false,"given":"Haobo","family":"Qi","sequence":"first","affiliation":[{"name":"School of Statistics, Beijing Normal University, Beijing 100875, China;"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-8260-2053","authenticated-orcid":false,"given":"Feifei","family":"Wang","sequence":"additional","affiliation":[{"name":"Center for Applied Statistics and School of Statistics, Renmin University of China, Beijing 100872, China;"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Hansheng","family":"Wang","sequence":"additional","affiliation":[{"name":"Guanghua School of Management, Peking University, Beijing 100871, China"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"109","reference":[{"key":"B1","doi-asserted-by":"crossref","unstructured":"Abadi M, Chu A, Goodfellow I, McMahan HB, Mironov I, Talwar K, Zhang L (2016) Deep learning with differential privacy.\n                      Proc. 2016 ACM SIGSAC Conf. Comput. Commun. Security CCS \u201816\n                      (Association for Computing Machinery, New York), 308\u2013318.","DOI":"10.1145\/2976749.2978318"},{"issue":"126","key":"B2","first-page":"1","volume":"17","author":"Arjevani Y","year":"2016","journal-title":"J. Machine Learn. Res."},{"key":"B3","unstructured":"Bach F, Moulines E (2013) Non-strongly-convex smooth stochastic approximation with convergence rate o(1\/n).\n                      Proc. 26th Internat. Conf. Neural Information Processing Systems NIPS\u201913\n                      , vol. 1 (Curran Associates Inc. Red Hook, NY), 773\u2013781."},{"key":"B4","doi-asserted-by":"publisher","DOI":"10.1016\/j.jeconom.2022.05.004"},{"key":"B5","volume-title":"Numerical Analysis","author":"Burden R","year":"2010","edition":"9"},{"key":"B6","doi-asserted-by":"crossref","unstructured":"Cai J, Liu Y, Liu X, Li J, Zhuang H (2023) Privacy-preserving federated cross-domain social recommendation.\n                      Trustworthy Federated Learn. First Internat. Workshop FL 2022 Held Conjunction IJCAI 2022 Revised Selected Papers\n                      (Springer International Publishing, New York), 144\u2013158.","DOI":"10.1007\/978-3-031-28996-5_11"},{"key":"B7","doi-asserted-by":"publisher","DOI":"10.1080\/01621459.2021.1891925"},{"key":"B8","doi-asserted-by":"publisher","DOI":"10.1214\/18-AOS1801"},{"key":"B9","doi-asserted-by":"publisher","DOI":"10.1145\/3606017"},{"key":"B10","doi-asserted-by":"publisher","DOI":"10.52202\/068431-0768"},{"key":"B11","unstructured":"Devlin J, Chang M-W, Lee K, Bert KT (2018) BERT: Pre-training of deep bidirectional transformers for language understanding. Preprint, submitted October 11, https:\/\/arxiv.org\/abs\/1810.04805v1."},{"key":"B12","doi-asserted-by":"publisher","DOI":"10.1111\/rssb.12454"},{"key":"B13","doi-asserted-by":"crossref","unstructured":"Dwork C, Lei J (2009) Differential privacy and robust statistics.\n                      Proc. Forty-First Annu. ACM Sympos. Theory Comput.\n                      (Association for Computing Machinery, New York), 371\u2013380,","DOI":"10.1145\/1536414.1536466"},{"key":"B14","doi-asserted-by":"crossref","unstructured":"Firdaus M, Larasati HT, Rhee K-H (2022) A secure federated learning framework using blockchain and differential privacy.\n                      2022 IEEE 9th Internat. Conf. Cyber Security Cloud Comput. CSCloud\n                      (IEEE, Piscataway, NJ), 18\u201323.","DOI":"10.1109\/CSCloud-EdgeCom54986.2022.00013"},{"key":"B15","unstructured":"Fu J, Hong Y, Ling X, Wang L, Ran X, Sun Z, Wang WH, Chen Z, Cao Y (2024) Differentially private federated learning: A systematic review. Preprint, submitted May 14, https:\/\/arxiv.org\/abs\/2405.08299."},{"key":"B16","first-page":"16937","volume-title":"Adv. Neural Inform. Processing Systems","volume":"33","author":"Geiping J","year":"2020"},{"key":"B17","unstructured":"Geyer RC, Klein T, Nabi M (2017) Differentially private federated learning: A client level perspective. Preprint, submitted December 20, https:\/\/arxiv.org\/abs\/1712.07557v1."},{"key":"B18","unstructured":"Gitman I, Lang H, Zhang P, Xiao L (2019) Understanding the role of momentum in stochastic gradient methods.\n                      Proc. 33rd Conf. Neural Inform. Processing Systems\n                      (Curran Associates, Inc., Red Hook, NY)."},{"key":"B19","doi-asserted-by":"publisher","DOI":"10.1109\/TPDS.2023.3244198"},{"key":"B20","doi-asserted-by":"crossref","unstructured":"Kim M, G\u00fcnl\u00fc O, Schaefer RF (2021) Federated learning with local differential privacy: Trade-offs between privacy, utility, and communication.\n                      ICASSP 2021\u20132021 IEEE Internat. Conf. Acoustics Speech Signal Processing ICASSP\n                      (IEEE, Piscataway, NJ), 2650\u20132654.","DOI":"10.1109\/ICASSP39728.2021.9413764"},{"key":"B21","doi-asserted-by":"publisher","DOI":"10.1109\/TPDS.2021.3138848"},{"key":"B22","unstructured":"Liu X, Kong W, Oh S (2022) Differential privacy and robust statistics in high dimensions.\n                      Proc. Thirty Fifth Conf. Learn. Theory,\n                      Proceedings of the Machine Learning Research, vol. 178 (PMLR, New York), 1167\u20131246."},{"key":"B23","unstructured":"Luo J, Xu Z (2023) Differential privacy statistical inference for a directed graph network model with covariates. Preprint, submitted December 8, https:\/\/arxiv.org\/abs\/2312.04903."},{"key":"B24","first-page":"1273","volume-title":"Artificial Intelligence and Statistics","author":"McMahan B","year":"2017"},{"key":"B25","doi-asserted-by":"crossref","unstructured":"Mironov I (2017) R\u00e9nyi differential privacy.\n                      2017 IEEE 30th Comput. Security Foundations Sympos. CSF\n                      (IEEE, Piscataway, NJ).","DOI":"10.1109\/CSF.2017.11"},{"key":"B26","unstructured":"Noble M, Bellet A, Dieuleveut A (2022) Differentially private federated learning on heterogeneous data.\n                      Internat. Conf. Artificial Intelligence Statist\n                      ., vol. 151 (PMLR, New York), 10110\u201310145."},{"key":"B27","volume-title":"Numerical Optimization","author":"Nocedal J","year":"2006","edition":"2"},{"key":"B28","doi-asserted-by":"crossref","unstructured":"Pichler G, Romanelli M, Vega LR, Piantanida P (2023) Perfectly accurate membership inference by a dishonest central server in federated learning.\n                      IEEE Trans. Dependable Secure Comput.\n                      21(4):4290\u20134296.","DOI":"10.1109\/TDSC.2023.3326230"},{"key":"B29","doi-asserted-by":"publisher","DOI":"10.1137\/0330046"},{"key":"B30","doi-asserted-by":"crossref","unstructured":"Qi H, Wang F, Wang H (2025) Privacy protection and statistical efficiency trade-off for federated learning. https:\/\/doi.org\/10.1287\/ijoc.2024.0554.cd, https:\/\/github.com\/INFORMSJoC\/2024.0554.","DOI":"10.1287\/ijoc.2024.0554.cd"},{"key":"B31","doi-asserted-by":"publisher","DOI":"10.1016\/j.neucom.2023.126407"},{"issue":"292","key":"B32","first-page":"1","volume":"23","author":"Wang Q","year":"2022","journal-title":"J. Machine Learn. Res."},{"key":"B33","doi-asserted-by":"publisher","DOI":"10.1109\/TNET.2024.3379439"},{"key":"B34","doi-asserted-by":"publisher","DOI":"10.1198\/jasa.2009.tm08651"},{"key":"B35","doi-asserted-by":"crossref","unstructured":"Watkins W, Wang H, Bae S, Tseng H-H, Cha J, Chen SY-C, Yoo S (2024) Quantum privacy aggregation of teacher ensembles (QPATE) for privacy-preserving quantum machine learning.\n                      ICASSP 2024\u20132024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)\n                      (IEEE, Piscataway, NJ), 6875\u20136879.","DOI":"10.1109\/ICASSP48485.2024.10447786"},{"key":"B36","doi-asserted-by":"publisher","DOI":"10.1109\/TIFS.2023.3293417"},{"key":"B37","doi-asserted-by":"publisher","DOI":"10.1109\/TIFS.2020.2988575"},{"key":"B38","unstructured":"Yousefpour A, Shilov I, Sablayrolles A, Testuggine D, Prasad K, Malek M, Nguyen J, et al. (2021) Opacus: User-friendly differential privacy library in PyTorch. Preprint, submitted September 25, https:\/\/arxiv.org\/abs\/2109.12298v1."},{"key":"B39","doi-asserted-by":"crossref","unstructured":"Zhang Q, Ma J, Lou J, Xiong L, Jiang X (2020) Towards training robust private aggregation of teacher ensembles under noisy labels.\n                      2020 IEEE Internat. Conf. Big Data Big Data\n                      (IEEE, Piscataway, NJ), 1103\u20131110,","DOI":"10.1109\/BigData50022.2020.9378234"},{"key":"B40","doi-asserted-by":"publisher","DOI":"10.1145\/3595185"},{"key":"B41","doi-asserted-by":"publisher","DOI":"10.1109\/TMLCN.2025.3528901"},{"key":"B42","unstructured":"Zheng Q, Chen S, Long Q, Su W (2021) Federated f-differential privacy.\n                      Internat. Conf. Artificial Intelligence Statist.\n                      (PMLR, New York), 2251\u20132259."}],"container-title":["INFORMS Journal on Computing"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/pubsonline.informs.org\/doi\/pdf\/10.1287\/ijoc.2024.0554","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,6,15]],"date-time":"2026-06-15T08:53:59Z","timestamp":1781513639000},"score":1,"resource":{"primary":{"URL":"https:\/\/pubsonline.informs.org\/doi\/10.1287\/ijoc.2024.0554"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,5]]},"references-count":42,"journal-issue":{"issue":"3","published-print":{"date-parts":[[2026,5]]}},"alternative-id":["10.1287\/ijoc.2024.0554"],"URL":"https:\/\/doi.org\/10.1287\/ijoc.2024.0554","relation":{},"ISSN":["1091-9856","1526-5528"],"issn-type":[{"value":"1091-9856","type":"print"},{"value":"1526-5528","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,5]]}}}