FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W4297629560> ?p ?o ?g. }

• W4297629560 abstract "Targeted training-set attacks inject malicious instances into the training set to cause a trained model to mislabel one or more specific test instances. This work proposes the task of target identification, which determines whether a specific test instance is the target of a training-set attack. Target identification can be combined with adversarial-instance identification to find (and remove) the attack instances, mitigating the attack with minimal impact on other predictions. Rather than focusing on a single attack method or data modality, we build on influence estimation, which quantifies each training instance's contribution to a model's prediction. We show that existing influence estimators' poor practical performance often derives from their over-reliance on training instances and iterations with large losses. Our renormalized influence estimators fix this weakness; they far outperform the original estimators at identifying influential groups of training examples in both adversarial and non-adversarial settings, even finding up to 100% of adversarial training instances with no clean-data false positives. Target identification then simplifies to detecting test instances with anomalous influence values. We demonstrate our method's effectiveness on backdoor and poisoning attacks across various data domains, including text, vision, and speech, as well as against a gray-box, adaptive attacker that specifically optimizes the adversarial instances to evade our method. Our source code is available at https://github.com/ZaydH/target_identification." @default.
• W4297629560 created "2022-09-30" @default.
• W4297629560 creator A5053373401 @default.
• W4297629560 creator A5072809662 @default.
• W4297629560 date "2022-11-07" @default.
• W4297629560 modified "2023-10-18" @default.
• W4297629560 title "Identifying a Training-Set Attack's Target Using Renormalized Influence Estimation" @default.
• W4297629560 cites W196871588 @default.
• W4297629560 cites W1976526581 @default.
• W4297629560 cites W2006903949 @default.
• W4297629560 cites W2030353609 @default.
• W4297629560 cites W2137130182 @default.
• W4297629560 cites W2498631646 @default.
• W4297629560 cites W2753783305 @default.
• W4297629560 cites W2807363941 @default.
• W4297629560 cites W2942091739 @default.
• W4297629560 cites W2964043980 @default.
• W4297629560 cites W2990270730 @default.
• W4297629560 cites W2995536569 @default.
• W4297629560 cites W3016970897 @default.
• W4297629560 cites W3022590799 @default.
• W4297629560 cites W3048759177 @default.
• W4297629560 cites W3093137868 @default.
• W4297629560 cites W3106646114 @default.
• W4297629560 cites W3116515605 @default.
• W4297629560 cites W3152758407 @default.
• W4297629560 cites W3170572542 @default.
• W4297629560 cites W3173692964 @default.
• W4297629560 cites W3199974954 @default.
• W4297629560 cites W3214399478 @default.
• W4297629560 cites W4214736395 @default.
• W4297629560 cites W4226301182 @default.
• W4297629560 cites W4229530126 @default.
• W4297629560 cites W4362131110 @default.
• W4297629560 doi "https://doi.org/10.1145/3548606.3559335" @default.
• W4297629560 hasPublicationYear "2022" @default.
• W4297629560 type Work @default.
• W4297629560 citedByCount "1" @default.
• W4297629560 countsByYear W42976295602023 @default.
• W4297629560 crossrefType "proceedings-article" @default.
• W4297629560 hasAuthorship W4297629560A5053373401 @default.
• W4297629560 hasAuthorship W4297629560A5072809662 @default.
• W4297629560 hasBestOaLocation W42976295602 @default.
• W4297629560 hasConcept C105795698 @default.
• W4297629560 hasConcept C116834253 @default.
• W4297629560 hasConcept C119857082 @default.
• W4297629560 hasConcept C121332964 @default.
• W4297629560 hasConcept C124101348 @default.
• W4297629560 hasConcept C127413603 @default.
• W4297629560 hasConcept C153294291 @default.
• W4297629560 hasConcept C154945302 @default.
• W4297629560 hasConcept C16910744 @default.
• W4297629560 hasConcept C169903167 @default.
• W4297629560 hasConcept C177264268 @default.
• W4297629560 hasConcept C185429906 @default.
• W4297629560 hasConcept C199360897 @default.
• W4297629560 hasConcept C201995342 @default.
• W4297629560 hasConcept C2777211547 @default.
• W4297629560 hasConcept C2780451532 @default.
• W4297629560 hasConcept C2781045450 @default.
• W4297629560 hasConcept C33923547 @default.
• W4297629560 hasConcept C37736160 @default.
• W4297629560 hasConcept C38652104 @default.
• W4297629560 hasConcept C41008148 @default.
• W4297629560 hasConcept C51632099 @default.
• W4297629560 hasConcept C59822182 @default.
• W4297629560 hasConcept C64869954 @default.
• W4297629560 hasConcept C65856478 @default.
• W4297629560 hasConcept C86803240 @default.
• W4297629560 hasConceptScore W4297629560C105795698 @default.
• W4297629560 hasConceptScore W4297629560C116834253 @default.
• W4297629560 hasConceptScore W4297629560C119857082 @default.
• W4297629560 hasConceptScore W4297629560C121332964 @default.
• W4297629560 hasConceptScore W4297629560C124101348 @default.
• W4297629560 hasConceptScore W4297629560C127413603 @default.
• W4297629560 hasConceptScore W4297629560C153294291 @default.
• W4297629560 hasConceptScore W4297629560C154945302 @default.
• W4297629560 hasConceptScore W4297629560C16910744 @default.
• W4297629560 hasConceptScore W4297629560C169903167 @default.
• W4297629560 hasConceptScore W4297629560C177264268 @default.
• W4297629560 hasConceptScore W4297629560C185429906 @default.
• W4297629560 hasConceptScore W4297629560C199360897 @default.
• W4297629560 hasConceptScore W4297629560C201995342 @default.
• W4297629560 hasConceptScore W4297629560C2777211547 @default.
• W4297629560 hasConceptScore W4297629560C2780451532 @default.
• W4297629560 hasConceptScore W4297629560C2781045450 @default.
• W4297629560 hasConceptScore W4297629560C33923547 @default.
• W4297629560 hasConceptScore W4297629560C37736160 @default.
• W4297629560 hasConceptScore W4297629560C38652104 @default.
• W4297629560 hasConceptScore W4297629560C41008148 @default.
• W4297629560 hasConceptScore W4297629560C51632099 @default.
• W4297629560 hasConceptScore W4297629560C59822182 @default.
• W4297629560 hasConceptScore W4297629560C64869954 @default.
• W4297629560 hasConceptScore W4297629560C65856478 @default.
• W4297629560 hasConceptScore W4297629560C86803240 @default.
• W4297629560 hasFunder F4320332180 @default.
• W4297629560 hasLocation W42976295601 @default.
• W4297629560 hasLocation W42976295602 @default.
• W4297629560 hasOpenAccess W4297629560 @default.
• W4297629560 hasPrimaryLocation W42976295601 @default.

• next