Matches in SemOpenAlex for { <https://semopenalex.org/work/W4375958086> ?p ?o ?g. }
Showing items 1 to 51 of
51
with 100 items per page.
- W4375958086 abstract "With the help of conditioning mechanisms, the state-of-the-art diffusion models have achieved tremendous success in guided image generation, particularly in text-to-image synthesis. To gain a better understanding of the training process and potential risks of text-to-image synthesis, we perform a systematic investigation of backdoor attack on text-to-image diffusion models and propose BadT2I, a general multimodal backdoor attack framework that tampers with image synthesis in diverse semantic levels. Specifically, we perform backdoor attacks on three levels of the vision semantics: Pixel-Backdoor, Object-Backdoor and Style-Backdoor. By utilizing a regularization loss, our methods efficiently inject backdoors into a large-scale text-to-image diffusion model while preserving its utility with benign inputs. We conduct empirical experiments on Stable Diffusion, the widely-used text-to-image diffusion model, demonstrating that the large-scale diffusion model can be easily backdoored within a few fine-tuning steps. We conduct additional experiments to explore the impact of different types of textual triggers. Besides, we discuss the backdoor persistence during further training, the findings of which provide insights for the development of backdoor defense methods." @default.
- W4375958086 created "2023-05-10" @default.
- W4375958086 creator A5008450480 @default.
- W4375958086 creator A5024922656 @default.
- W4375958086 creator A5032734653 @default.
- W4375958086 creator A5035938543 @default.
- W4375958086 creator A5068755794 @default.
- W4375958086 creator A5088688674 @default.
- W4375958086 date "2023-05-06" @default.
- W4375958086 modified "2023-09-27" @default.
- W4375958086 title "Text-to-Image Diffusion Models can be Easily Backdoored through Multimodal Data Poisoning" @default.
- W4375958086 doi "https://doi.org/10.48550/arxiv.2305.04175" @default.
- W4375958086 hasPublicationYear "2023" @default.
- W4375958086 type Work @default.
- W4375958086 citedByCount "0" @default.
- W4375958086 crossrefType "posted-content" @default.
- W4375958086 hasAuthorship W4375958086A5008450480 @default.
- W4375958086 hasAuthorship W4375958086A5024922656 @default.
- W4375958086 hasAuthorship W4375958086A5032734653 @default.
- W4375958086 hasAuthorship W4375958086A5035938543 @default.
- W4375958086 hasAuthorship W4375958086A5068755794 @default.
- W4375958086 hasAuthorship W4375958086A5088688674 @default.
- W4375958086 hasBestOaLocation W43759580861 @default.
- W4375958086 hasConcept C115961682 @default.
- W4375958086 hasConcept C154945302 @default.
- W4375958086 hasConcept C2781045450 @default.
- W4375958086 hasConcept C38652104 @default.
- W4375958086 hasConcept C41008148 @default.
- W4375958086 hasConcept C80444323 @default.
- W4375958086 hasConceptScore W4375958086C115961682 @default.
- W4375958086 hasConceptScore W4375958086C154945302 @default.
- W4375958086 hasConceptScore W4375958086C2781045450 @default.
- W4375958086 hasConceptScore W4375958086C38652104 @default.
- W4375958086 hasConceptScore W4375958086C41008148 @default.
- W4375958086 hasConceptScore W4375958086C80444323 @default.
- W4375958086 hasLocation W43759580861 @default.
- W4375958086 hasOpenAccess W4375958086 @default.
- W4375958086 hasPrimaryLocation W43759580861 @default.
- W4375958086 hasRelatedWork W2795359650 @default.
- W4375958086 hasRelatedWork W3107474891 @default.
- W4375958086 hasRelatedWork W3213537515 @default.
- W4375958086 hasRelatedWork W4292171917 @default.
- W4375958086 hasRelatedWork W4297741810 @default.
- W4375958086 hasRelatedWork W4308504118 @default.
- W4375958086 hasRelatedWork W4324297007 @default.
- W4375958086 hasRelatedWork W4360818952 @default.
- W4375958086 hasRelatedWork W4361230562 @default.
- W4375958086 hasRelatedWork W4362471971 @default.
- W4375958086 isParatext "false" @default.
- W4375958086 isRetracted "false" @default.
- W4375958086 workType "article" @default.