English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Federated learning (FL) provides an efficient paradigm to jointly train aglobal model leveraging data from distributed users. As local training datacomes from different users who may not be trustworthy, several studies haveshown that FL is vulnerable to poisoning attacks. Meanwhile, to protect theprivacy of local users, FL is usually trained in a differentially private way(DPFL). Thus, in this paper, we ask: What are the underlying connectionsbetween differential privacy and certified robustness in FL against poisoningattacks? Can we leverage the innate privacy property of DPFL to providecertified robustness for FL? Can we further improve the privacy of FL toimprove such robustness certification? We first investigate both user-level andinstance-level privacy of FL and provide formal privacy analysis to achieveimproved instance-level privacy. We then provide two robustness certificationcriteria: certified prediction and certified attack inefficacy for DPFL on bothuser and instance levels. Theoretically, we provide the certified robustness ofDPFL based on both criteria given a bounded number of adversarial users orinstances. Empirically, we conduct extensive experiments to verify our theoriesunder a range of poisoning attacks on different datasets. We find thatincreasing the level of privacy protection in DPFL results in strongercertified attack inefficacy; however, it does not necessarily lead to astronger certified prediction. Thus, achieving the optimal certified predictionrequires a proper balance between privacy and utility loss.

Unraveling the Connections between Privacy and Certified Robustness in Federated Learning Against Poisoning Attacks