Harnessing Hyperbolic Geometry for Harmful Prompt Detection and Sanitization

Computer Science > Cryptography and Security [Submitted on 7 Apr 2026] Harnessing Hyperbolic Geometry for Harmful Prompt Detection and Sanitization Igor Maljkovic, Maria Rosaria Briglia, Iacopo Masi, Antonio Emanuele Cinà, Fabio Roli Vision-Language Models (VLMs) have become essential for tasks such as image synthesis, captioning, and retrieval by aligning textual and visual information in a shared embedding space. Yet, this flexibility also makes them vulnerable to malicious prompts designed to produce unsafe content, raising critical safety concerns. Existing defenses either rely on blacklist filters, which are easily circumvented, or on heavy classifier-based systems, both of which are costly and fragile under embedding-level attacks. We address these challenges with two complementary components: Hyperbolic Prompt Espial (HyPE) and Hyperbolic Prompt Sanitization (HyPS). HyPE is a lightweight anomaly detector that leverages the structured geometry of hyperbolic space to model benign prompts and detect harmful ones as outliers. HyPS builds on this detection by applying explainable attribution methods to identify and selectively modify harmful words, neutralizing unsafe intent while preserving the original semantics of user prompts. Through extensive experiments across multiple datasets and adversarial scenarios, we prove that our framework consistently outperforms prior defenses in both detection accuracy and robustness. Together, HyPE and HyPS offer an efficient, interpretable, and resilient approach to safeguarding VLMs against malicious prompt misuse. Comments: Paper accepted at ICLR 2026. Webpage available at: this https URL Subjects: Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV) Cite as: arXiv:2604.06285 [cs.CR]   (or arXiv:2604.06285v1 [cs.CR] for this version)   https://doi.org/10.48550/arXiv.2604.06285 Focus to learn more Submission history From: Antonio Emanuele Cinà [view email] [v1] Tue, 7 Apr 2026 12:49:24 UTC (18,634 KB) Access Paper: HTML (experimental) view license Current browse context: cs.CR < prev   |   next > new | recent | 2026-04 Change to browse by: cs cs.AI cs.CV References & Citations NASA ADS Google Scholar Semantic Scholar Export BibTeX Citation Bookmark Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer (What is the Explorer?) Connected Papers Toggle Connected Papers (What is Connected Papers?) Litmaps Toggle Litmaps (What is Litmaps?) scite.ai Toggle scite Smart Citations (What are Smart Citations?) Code, Data, Media Demos Related Papers About arXivLabs Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)