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

Data heterogeneity has been identified as one of the key features infederated learning but often overlooked in the lens of robustness toadversarial attacks. This paper focuses on characterizing and understanding itsimpact on backdooring attacks in federated learning through comprehensiveexperiments using synthetic and the LEAF benchmarks. The initial impressiondriven by our experimental results suggests that data heterogeneity is thedominant factor in the effectiveness of attacks and it may be a redemption fordefending against backdooring as it makes the attack less efficient, morechallenging to design effective attack strategies, and the attack result alsobecomes less predictable. However, with further investigations, we found dataheterogeneity is more of a curse than a redemption as the attack effectivenesscan be significantly boosted by simply adjusting the client-side backdooringtiming. More importantly,data heterogeneity may result in overfitting at thelocal training of benign clients, which can be utilized by attackers todisguise themselves and fool skewed-feature based defenses. In addition,effective attack strategies can be made by adjusting attack data distribution.Finally, we discuss the potential directions of defending the curses brought bydata heterogeneity. The results and lessons learned from our extensiveexperiments and analysis offer new insights for designing robust federatedlearning methods and systems

Curse or Redemption? How Data Heterogeneity Affects the Robustness of Federated Learning