InfoSeverity comes from the source advisory when it gives one. Otherwise the classifier assigns it from the source text.ResearchPeer-reviewedLLM-specificPrivacy-preserving for user-uploaded images and text in Vision-Language ModelsSourceElsevier Security Journals(opens in a new tab)PublishedApril 28, 2026Record updatedOctober 9, 2026View JSONCopy API URLCopy IEEE citationShare on XShare on LinkedInCopy linkTopicsMultimodal models