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/

Zendy Plus

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

Zendy Tools

Zendy Open

Federated Learning (FL) enables multiple parties to train machine learningmodels collaboratively without sharing the raw training data. However, thefederated nature of FL enables malicious clients to influence a trained modelby injecting error model updates via Byzantine or backdoor attacks. To detectmalicious model updates, a typical approach is to measure the distance betweeneach model update and a \textit{ground-truth model update}. To find such\textit{ground-truth model updates}, existing defenses either require a benignroot dataset on the server (e.g., FLTrust) or simply use trimmed mean or medianas the threshold for clipping (e.g., FLAME). However, such benign root datasetsare impractical, and the trimmed mean or median may also eliminatecontributions from these underrepresented datasets.  In this paper, we propose a generic solution, namely FedTruth, to defendagainst model poisoning attacks in FL, where the \textit{ground-truth modelupdate} (i.e., the global model update) will be estimated among all the modelupdates with dynamic aggregation weights. Specifically, FedTruth does not havespecific assumptions on the benign or malicious data distribution or access toa benign root dataset. Moreover, FedTruth considers the potential contributionsfrom all benign clients. Our empirical results show that FedTruth can reducethe impacts of poisoned model updates against both Byzantine and backdoorattacks, and is also efficient in large-scale FL systems.

Identifying the Truth of Global Model: A Generic Solution to Defend Against Byzantine and Backdoor Attacks in Federated Learning (full version)