Matches in SemOpenAlex for { <https://semopenalex.org/work/W3035237306> ?p ?o ?g. }
Showing items 1 to 84 of
84
with 100 items per page.
- W3035237306 abstract "With the rapid growth of visual content, deep learning to hash is gaining popularity in the image retrieval community recently. Although it greatly facilitates search efficiency, privacy is also at risks when images on the web are retrieved at a large scale and exploited as a rich mine of personal information. An adversary can extract private images by querying similar images from the targeted category for any usable model. Existing methods based on image processing preserve privacy at a sacrifice of perceptual quality. In this paper, we propose a new mechanism based on adversarial examples to stash'' private images in the deep hash space while maintaining perceptual similarity. We first find that a simple approach of hamming distance maximization is not robust against brute-force adversaries. Then we develop a new loss function by maximizing the hamming distance to not only the original category, but also the centers from all the classes, partitioned into clusters of various sizes. The extensive experiment shows that the proposed defense can harden the attacker's efforts by 2-7 orders of magnitude, without significant increase of computational overhead and perceptual degradation. We also demonstrate 30-60% transferability in hash space with a black-box setting. The code is available at: https://github.com/sugarruy/hashstash" @default.
- W3035237306 created "2020-06-19" @default.
- W3035237306 creator A5007601438 @default.
- W3035237306 creator A5011508278 @default.
- W3035237306 creator A5046078938 @default.
- W3035237306 date "2020-06-01" @default.
- W3035237306 modified "2023-10-14" @default.
- W3035237306 title "Evade Deep Image Retrieval by Stashing Private Images in the Hash Space" @default.
- W3035237306 cites W1566135517 @default.
- W3035237306 cites W1913628733 @default.
- W3035237306 cites W1972524097 @default.
- W3035237306 cites W2026254891 @default.
- W3035237306 cites W2069169703 @default.
- W3035237306 cites W2108598243 @default.
- W3035237306 cites W2124199175 @default.
- W3035237306 cites W2133665775 @default.
- W3035237306 cites W2148794952 @default.
- W3035237306 cites W2161969291 @default.
- W3035237306 cites W2285252196 @default.
- W3035237306 cites W2392997304 @default.
- W3035237306 cites W2411707397 @default.
- W3035237306 cites W2464915613 @default.
- W3035237306 cites W2488334223 @default.
- W3035237306 cites W2508120068 @default.
- W3035237306 cites W2603766943 @default.
- W3035237306 cites W2770975812 @default.
- W3035237306 cites W2963857521 @default.
- W3035237306 cites W2964280870 @default.
- W3035237306 cites W3105655001 @default.
- W3035237306 doi "https://doi.org/10.1109/cvpr42600.2020.00967" @default.
- W3035237306 hasPublicationYear "2020" @default.
- W3035237306 type Work @default.
- W3035237306 sameAs 3035237306 @default.
- W3035237306 citedByCount "8" @default.
- W3035237306 countsByYear W30352373062021 @default.
- W3035237306 countsByYear W30352373062022 @default.
- W3035237306 countsByYear W30352373062023 @default.
- W3035237306 crossrefType "proceedings-article" @default.
- W3035237306 hasAuthorship W3035237306A5007601438 @default.
- W3035237306 hasAuthorship W3035237306A5011508278 @default.
- W3035237306 hasAuthorship W3035237306A5046078938 @default.
- W3035237306 hasBestOaLocation W30352373062 @default.
- W3035237306 hasConcept C11413529 @default.
- W3035237306 hasConcept C115961682 @default.
- W3035237306 hasConcept C154945302 @default.
- W3035237306 hasConcept C157125643 @default.
- W3035237306 hasConcept C1667742 @default.
- W3035237306 hasConcept C2779494224 @default.
- W3035237306 hasConcept C38652104 @default.
- W3035237306 hasConcept C41008148 @default.
- W3035237306 hasConcept C57273362 @default.
- W3035237306 hasConcept C73150493 @default.
- W3035237306 hasConcept C80444323 @default.
- W3035237306 hasConcept C99138194 @default.
- W3035237306 hasConceptScore W3035237306C11413529 @default.
- W3035237306 hasConceptScore W3035237306C115961682 @default.
- W3035237306 hasConceptScore W3035237306C154945302 @default.
- W3035237306 hasConceptScore W3035237306C157125643 @default.
- W3035237306 hasConceptScore W3035237306C1667742 @default.
- W3035237306 hasConceptScore W3035237306C2779494224 @default.
- W3035237306 hasConceptScore W3035237306C38652104 @default.
- W3035237306 hasConceptScore W3035237306C41008148 @default.
- W3035237306 hasConceptScore W3035237306C57273362 @default.
- W3035237306 hasConceptScore W3035237306C73150493 @default.
- W3035237306 hasConceptScore W3035237306C80444323 @default.
- W3035237306 hasConceptScore W3035237306C99138194 @default.
- W3035237306 hasLocation W30352373061 @default.
- W3035237306 hasLocation W30352373062 @default.
- W3035237306 hasOpenAccess W3035237306 @default.
- W3035237306 hasPrimaryLocation W30352373061 @default.
- W3035237306 hasRelatedWork W2159024673 @default.
- W3035237306 hasRelatedWork W2574678123 @default.
- W3035237306 hasRelatedWork W2894874917 @default.
- W3035237306 hasRelatedWork W2968165261 @default.
- W3035237306 hasRelatedWork W2981938443 @default.
- W3035237306 hasRelatedWork W3016242994 @default.
- W3035237306 hasRelatedWork W3023219121 @default.
- W3035237306 hasRelatedWork W3210945214 @default.
- W3035237306 hasRelatedWork W4287817783 @default.
- W3035237306 hasRelatedWork W4288718166 @default.
- W3035237306 isParatext "false" @default.
- W3035237306 isRetracted "false" @default.
- W3035237306 magId "3035237306" @default.
- W3035237306 workType "article" @default.