Matches in SemOpenAlex for { <https://semopenalex.org/work/W4285555619> ?p ?o ?g. }
Showing items 1 to 94 of
94
with 100 items per page.
- W4285555619 endingPage "316" @default.
- W4285555619 startingPage "303" @default.
- W4285555619 abstract "Gradient leakage attacks are considered one of the wickedest privacy threats in deep learning as attackers covertly spy gradient updates during iterative training without compromising model training quality, and yet secretly reconstruct sensitive training data using leaked gradients with high attack success rate. Although deep learning with differential privacy is a defacto standard for publishing deep learning models with differential privacy guarantee, we show that differentially private algorithms with fixed privacy parameters are vulnerable against gradient leakage attacks. This paper investigates alternative approaches to gradient leakage resilient deep learning with differential privacy (DP). <i>First</i>, we analyze existing implementation of deep learning with differential privacy, which use fixed noise variance to injects constant noise to the gradients in all layers using fixed privacy parameters. Despite the DP guarantee provided, the method suffers from low accuracy and is vulnerable to gradient leakage attacks. <i>Second</i>, we present a gradient leakage resilient deep learning approach with differential privacy guarantee by using dynamic privacy parameters. Unlike fixed-parameter strategies that result in constant noise variance, different dynamic parameter strategies present alternative techniques to introduce adaptive noise variance and adaptive noise injection which are closely aligned to the trend of gradient updates during differentially private model training. <i>Finally</i>, we describe four complementary metrics to evaluate and compare alternative approaches. Extensive experiments on six benchmark datasets show that differentially private deep learning with dynamic privacy parameters outperforms the deep learning using fixed DP parameters, and existing adaptive clipping approaches in all aspects: compelling accuracy performance, strong differential privacy guarantee, and high attack resilience." @default.
- W4285555619 created "2022-07-16" @default.
- W4285555619 creator A5017360203 @default.
- W4285555619 creator A5069331320 @default.
- W4285555619 date "2022-01-01" @default.
- W4285555619 modified "2023-10-17" @default.
- W4285555619 title "Gradient Leakage Attack Resilient Deep Learning" @default.
- W4285555619 cites W1992926795 @default.
- W4285555619 cites W2051267297 @default.
- W4285555619 cites W2101771965 @default.
- W4285555619 cites W2167372639 @default.
- W4285555619 cites W2180612164 @default.
- W4285555619 cites W2473418344 @default.
- W4285555619 cites W2535690855 @default.
- W4285555619 cites W2591882872 @default.
- W4285555619 cites W2757528734 @default.
- W4285555619 cites W2781091734 @default.
- W4285555619 cites W2897830718 @default.
- W4285555619 cites W2919337578 @default.
- W4285555619 cites W2963313259 @default.
- W4285555619 cites W2963456518 @default.
- W4285555619 cites W2963857521 @default.
- W4285555619 cites W2964054038 @default.
- W4285555619 cites W2964162474 @default.
- W4285555619 cites W2970606380 @default.
- W4285555619 cites W3047250731 @default.
- W4285555619 cites W3085804918 @default.
- W4285555619 cites W3102360395 @default.
- W4285555619 cites W3138758728 @default.
- W4285555619 cites W3161357657 @default.
- W4285555619 cites W3175192640 @default.
- W4285555619 cites W4205228770 @default.
- W4285555619 doi "https://doi.org/10.1109/tifs.2021.3139777" @default.
- W4285555619 hasPublicationYear "2022" @default.
- W4285555619 type Work @default.
- W4285555619 citedByCount "6" @default.
- W4285555619 countsByYear W42855556192023 @default.
- W4285555619 crossrefType "journal-article" @default.
- W4285555619 hasAuthorship W4285555619A5017360203 @default.
- W4285555619 hasAuthorship W4285555619A5069331320 @default.
- W4285555619 hasBestOaLocation W42855556192 @default.
- W4285555619 hasConcept C108583219 @default.
- W4285555619 hasConcept C11413529 @default.
- W4285555619 hasConcept C115961682 @default.
- W4285555619 hasConcept C119857082 @default.
- W4285555619 hasConcept C13280743 @default.
- W4285555619 hasConcept C139719470 @default.
- W4285555619 hasConcept C154945302 @default.
- W4285555619 hasConcept C162324750 @default.
- W4285555619 hasConcept C185798385 @default.
- W4285555619 hasConcept C205649164 @default.
- W4285555619 hasConcept C23130292 @default.
- W4285555619 hasConcept C2777042071 @default.
- W4285555619 hasConcept C2779201187 @default.
- W4285555619 hasConcept C38652104 @default.
- W4285555619 hasConcept C41008148 @default.
- W4285555619 hasConcept C99498987 @default.
- W4285555619 hasConceptScore W4285555619C108583219 @default.
- W4285555619 hasConceptScore W4285555619C11413529 @default.
- W4285555619 hasConceptScore W4285555619C115961682 @default.
- W4285555619 hasConceptScore W4285555619C119857082 @default.
- W4285555619 hasConceptScore W4285555619C13280743 @default.
- W4285555619 hasConceptScore W4285555619C139719470 @default.
- W4285555619 hasConceptScore W4285555619C154945302 @default.
- W4285555619 hasConceptScore W4285555619C162324750 @default.
- W4285555619 hasConceptScore W4285555619C185798385 @default.
- W4285555619 hasConceptScore W4285555619C205649164 @default.
- W4285555619 hasConceptScore W4285555619C23130292 @default.
- W4285555619 hasConceptScore W4285555619C2777042071 @default.
- W4285555619 hasConceptScore W4285555619C2779201187 @default.
- W4285555619 hasConceptScore W4285555619C38652104 @default.
- W4285555619 hasConceptScore W4285555619C41008148 @default.
- W4285555619 hasConceptScore W4285555619C99498987 @default.
- W4285555619 hasFunder F4320306076 @default.
- W4285555619 hasLocation W42855556191 @default.
- W4285555619 hasLocation W42855556192 @default.
- W4285555619 hasOpenAccess W4285555619 @default.
- W4285555619 hasPrimaryLocation W42855556191 @default.
- W4285555619 hasRelatedWork W2795261237 @default.
- W4285555619 hasRelatedWork W3014300295 @default.
- W4285555619 hasRelatedWork W3164822677 @default.
- W4285555619 hasRelatedWork W4223943233 @default.
- W4285555619 hasRelatedWork W4225161397 @default.
- W4285555619 hasRelatedWork W4312200629 @default.
- W4285555619 hasRelatedWork W4360585206 @default.
- W4285555619 hasRelatedWork W4364306694 @default.
- W4285555619 hasRelatedWork W4380075502 @default.
- W4285555619 hasRelatedWork W4380086463 @default.
- W4285555619 hasVolume "17" @default.
- W4285555619 isParatext "false" @default.
- W4285555619 isRetracted "false" @default.
- W4285555619 workType "article" @default.