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

Gradient-based training in federated learning is known to be vulnerable tofaulty/malicious clients, which are often modeled as Byzantine clients. To thisend, previous work either makes use of auxiliary data at parameter server toverify the received gradients (e.g., by computing validation error rate) orleverages statistic-based methods (e.g. median and Krum) to identify and removemalicious gradients from Byzantine clients. In this paper, we remark thatauxiliary data may not always be available in practice and focus on thestatistic-based approach. However, recent work on model poisoning attacks hasshown that well-crafted attacks can circumvent most of median- anddistance-based statistical defense methods, making malicious gradientsindistinguishable from honest ones. To tackle this challenge, we show that theelement-wise sign of gradient vector can provide valuable insight in detectingmodel poisoning attacks. Based on our theoretical analysis of the\textit{Little is Enough} attack, we propose a novel approach called\textit{SignGuard} to enable Byzantine-robust federated learning throughcollaborative malicious gradient filtering. More precisely, the receivedgradients are first processed to generate relevant magnitude, sign, andsimilarity statistics, which are then collaboratively utilized by multiplefilters to eliminate malicious gradients before final aggregation. Finally,extensive experiments of image and text classification tasks are conductedunder recently proposed attacks and defense strategies. The numerical resultsdemonstrate the effectiveness and superiority of our proposed approach. Thecode is available at \textit{\url{https://github.com/JianXu95/SignGuard}}

Byzantine-robust Federated Learning through Collaborative Malicious Gradient Filtering