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

As a distributed machine learning paradigm, Federated Learning (FL) enableslarge-scale clients to collaboratively train a model without sharing their rawdata. However, due to the lack of data auditing for untrusted clients, FL isvulnerable to poisoning attacks, especially backdoor attacks. By using poisoneddata for local training or directly changing the model parameters, attackerscan easily inject backdoors into the model, which can trigger the model to makemisclassification of targeted patterns in images. To address these issues, wepropose a novel data-free trigger-generation-based defense approach based onthe two characteristics of backdoor attacks: i) triggers are learned fasterthan normal knowledge, and ii) trigger patterns have a greater effect on imageclassification than normal class patterns. Our approach generates the imageswith newly learned knowledge by identifying the differences between the old andnew global models, and filters trigger images by evaluating the effect of thesegenerated images. By using these trigger images, our approach eliminatespoisoned models to ensure the updated global model is benign. Comprehensiveexperiments demonstrate that our approach can defend against almost all theexisting types of backdoor attacks and outperform all the sevenstate-of-the-art defense methods with both IID and non-IID scenarios.Especially, our approach can successfully defend against the backdoor attackeven when 80\% of the clients are malicious.

Protect Federated Learning Against Backdoor Attacks via Data-Free Trigger Generation