Matches in SemOpenAlex for { <https://semopenalex.org/work/W4311457580> ?p ?o ?g. }
Showing items 1 to 82 of
82
with 100 items per page.
- W4311457580 endingPage "110178" @default.
- W4311457580 startingPage "110178" @default.
- W4311457580 abstract "Federated learning (FL) enables learning a global machine learning model from data distributed among a set of participating workers. This makes it possible (i) to train more accurate models due to learning from rich, joint training data and (ii) to improve privacy by not sharing the workers’ local private data with others. However, the distributed nature of FL makes it vulnerable to targeted poisoning attacks that negatively impact on the integrity of the learned model while, unfortunately, being difficult to detect. Existing defenses against those attacks are limited by assumptions on the workers’ data distribution and/or are ill-suited to high-dimensional models. In this paper, we analyze targeted attacks against FL, specifically label-flipping and backdoor attacks, and find that the neurons in the last layer of a deep learning (DL) model that are related to these attacks exhibit a different behavior from the unrelated neurons. This makes the last-layer gradients valuable features for attack detection. Accordingly, we propose FL-Defender to combat FL targeted attacks. It consists of (i) engineering robust discriminative features by calculating the worker-wise angle similarity for the workers’ last-layer gradients, (ii) compressing the resulting similarity vectors using PCA to reduce redundant information, and (iii) re-weighting the workers’ updates based on their deviation from the centroid of the compressed similarity vectors. Experiments on three data sets show the effectiveness of our method in defending against label-flipping and backdoor attacks. Compared to several state-of-the-art defenses, FL-Defender achieves the lowest attack success rates while maintaining the main task accuracy." @default.
- W4311457580 created "2022-12-26" @default.
- W4311457580 creator A5016893172 @default.
- W4311457580 creator A5077470825 @default.
- W4311457580 date "2023-01-01" @default.
- W4311457580 modified "2023-09-30" @default.
- W4311457580 title "FL-Defender: Combating targeted attacks in federated learning" @default.
- W4311457580 cites W2089468765 @default.
- W4311457580 cites W2144643813 @default.
- W4311457580 cites W2194775991 @default.
- W4311457580 cites W2962804345 @default.
- W4311457580 cites W3046449784 @default.
- W4311457580 cites W3138513656 @default.
- W4311457580 cites W3180515657 @default.
- W4311457580 cites W3200969281 @default.
- W4311457580 cites W3203932036 @default.
- W4311457580 cites W4306830536 @default.
- W4311457580 doi "https://doi.org/10.1016/j.knosys.2022.110178" @default.
- W4311457580 hasPublicationYear "2023" @default.
- W4311457580 type Work @default.
- W4311457580 citedByCount "3" @default.
- W4311457580 countsByYear W43114575802023 @default.
- W4311457580 crossrefType "journal-article" @default.
- W4311457580 hasAuthorship W4311457580A5016893172 @default.
- W4311457580 hasAuthorship W4311457580A5077470825 @default.
- W4311457580 hasBestOaLocation W43114575802 @default.
- W4311457580 hasConcept C103278499 @default.
- W4311457580 hasConcept C108583219 @default.
- W4311457580 hasConcept C115961682 @default.
- W4311457580 hasConcept C119857082 @default.
- W4311457580 hasConcept C124101348 @default.
- W4311457580 hasConcept C126838900 @default.
- W4311457580 hasConcept C154945302 @default.
- W4311457580 hasConcept C177264268 @default.
- W4311457580 hasConcept C178790620 @default.
- W4311457580 hasConcept C183115368 @default.
- W4311457580 hasConcept C185592680 @default.
- W4311457580 hasConcept C199360897 @default.
- W4311457580 hasConcept C2779227376 @default.
- W4311457580 hasConcept C2781045450 @default.
- W4311457580 hasConcept C38652104 @default.
- W4311457580 hasConcept C41008148 @default.
- W4311457580 hasConcept C71924100 @default.
- W4311457580 hasConcept C97931131 @default.
- W4311457580 hasConceptScore W4311457580C103278499 @default.
- W4311457580 hasConceptScore W4311457580C108583219 @default.
- W4311457580 hasConceptScore W4311457580C115961682 @default.
- W4311457580 hasConceptScore W4311457580C119857082 @default.
- W4311457580 hasConceptScore W4311457580C124101348 @default.
- W4311457580 hasConceptScore W4311457580C126838900 @default.
- W4311457580 hasConceptScore W4311457580C154945302 @default.
- W4311457580 hasConceptScore W4311457580C177264268 @default.
- W4311457580 hasConceptScore W4311457580C178790620 @default.
- W4311457580 hasConceptScore W4311457580C183115368 @default.
- W4311457580 hasConceptScore W4311457580C185592680 @default.
- W4311457580 hasConceptScore W4311457580C199360897 @default.
- W4311457580 hasConceptScore W4311457580C2779227376 @default.
- W4311457580 hasConceptScore W4311457580C2781045450 @default.
- W4311457580 hasConceptScore W4311457580C38652104 @default.
- W4311457580 hasConceptScore W4311457580C41008148 @default.
- W4311457580 hasConceptScore W4311457580C71924100 @default.
- W4311457580 hasConceptScore W4311457580C97931131 @default.
- W4311457580 hasLocation W43114575801 @default.
- W4311457580 hasLocation W43114575802 @default.
- W4311457580 hasOpenAccess W4311457580 @default.
- W4311457580 hasPrimaryLocation W43114575801 @default.
- W4311457580 hasRelatedWork W2353457699 @default.
- W4311457580 hasRelatedWork W3014300295 @default.
- W4311457580 hasRelatedWork W3164822677 @default.
- W4311457580 hasRelatedWork W4223943233 @default.
- W4311457580 hasRelatedWork W4225161397 @default.
- W4311457580 hasRelatedWork W4309045103 @default.
- W4311457580 hasRelatedWork W4312200629 @default.
- W4311457580 hasRelatedWork W4319994054 @default.
- W4311457580 hasRelatedWork W4360585206 @default.
- W4311457580 hasRelatedWork W4364306694 @default.
- W4311457580 hasVolume "260" @default.
- W4311457580 isParatext "false" @default.
- W4311457580 isRetracted "false" @default.
- W4311457580 workType "article" @default.