{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,14]],"date-time":"2026-07-14T07:33:37Z","timestamp":1784014417386,"version":"3.55.0"},"reference-count":61,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","license":[{"start":{"date-parts":[[2025,1,1]],"date-time":"2025-01-01T00:00:00Z","timestamp":1735689600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2025,1,1]],"date-time":"2025-01-01T00:00:00Z","timestamp":1735689600000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2025,1,1]],"date-time":"2025-01-01T00:00:00Z","timestamp":1735689600000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"publisher","award":["62206009"],"award-info":[{"award-number":["62206009"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]},{"name":"State Key Laboratory of Complex and Critical Software Environment (CCSE), Aeronautical Science Fund","award":["20230017051001"],"award-info":[{"award-number":["20230017051001"]}]},{"DOI":"10.13039\/501100001809","name":"Fundamental Research Funds for the Central Universities","doi-asserted-by":"publisher","id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE Trans.Inform.Forensic Secur."],"published-print":{"date-parts":[[2025]]},"DOI":"10.1109\/tifs.2025.3583249","type":"journal-article","created":{"date-parts":[[2025,6,30]],"date-time":"2025-06-30T13:37:31Z","timestamp":1751290651000},"page":"7153-7165","source":"Crossref","is-referenced-by-count":25,"title":["Jailbreak Vision Language Models via Bi-Modal Adversarial Prompt"],"prefix":"10.1109","volume":"20","author":[{"ORCID":"https:\/\/orcid.org\/0009-0007-7393-7362","authenticated-orcid":false,"given":"Zonghao","family":"Ying","sequence":"first","affiliation":[{"name":"SKLCCSE, School of Artificial Intelligence, Beihang University, Beijing, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-4224-1318","authenticated-orcid":false,"given":"Aishan","family":"Liu","sequence":"additional","affiliation":[{"name":"SKLCCSE, School of Computer Science and Engineering, Beihang University, Beijing, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9874-6828","authenticated-orcid":false,"given":"Tianyuan","family":"Zhang","sequence":"additional","affiliation":[{"name":"SKLCCSE, School of Computer Science and Engineering, Beihang University, Beijing, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Zhengmin","family":"Yu","sequence":"additional","affiliation":[{"name":"School of Computer Science, Fudan University, Shanghai, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-6154-0233","authenticated-orcid":false,"given":"Siyuan","family":"Liang","sequence":"additional","affiliation":[{"name":"School of Computing, National University of Singapore, Queenstown, Singapore"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-8425-4195","authenticated-orcid":false,"given":"Xianglong","family":"Liu","sequence":"additional","affiliation":[{"name":"SKLCCSE, School of Computer Science and Engineering, Beihang University, Beijing, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-7225-5449","authenticated-orcid":false,"given":"Dacheng","family":"Tao","sequence":"additional","affiliation":[{"name":"College of Computing and Data Science, Nanyang Technological University, Jurong West, Singapore"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","first-page":"34892","article-title":"Visual instruction tuning","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Liu"},{"key":"ref2","volume-title":"Gpt-4o","year":"2024"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01750"},{"key":"ref4","article-title":"Visual question answering instruction: Unlocking multimodal large language model to domain-specific visual multitasks","author":"Lee","year":"2024","journal-title":"arXiv:2402.08360"},{"key":"ref5","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV48922.2021.00213"},{"key":"ref6","article-title":"An image is worth 1000 lies: Adversarial transferability across prompts on vision-language models","author":"Luo","year":"2024","journal-title":"arXiv:2403.09766"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1109\/ICCVW60793.2023.00395"},{"key":"ref8","article-title":"On the robustness of large multimodal models against image adversarial attacks","author":"Cui","year":"2023","journal-title":"arXiv:2312.03777"},{"key":"ref9","first-page":"1","article-title":"Jailbreak in pieces: Compositional adversarial attacks on multi-modal language models","volume-title":"Proc. Int. Conf. Learn. Represent","author":"Shayegani"},{"key":"ref10","article-title":"GPTFUZZER: Red teaming large language models with auto-generated jailbreak prompts","author":"Yu","year":"2023","journal-title":"arXiv:2309.10253"},{"key":"ref11","first-page":"1","article-title":"Jailbroken: How does LLM safety training fail?","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Wei"},{"key":"ref12","article-title":"Jailbreaking black box large language models in twenty queries","author":"Chao","year":"2023","journal-title":"arXiv:2310.08419"},{"key":"ref13","first-page":"61478","article-title":"Are aligned neural networks adversarially aligned?","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Carlini"},{"key":"ref14","first-page":"55006","article-title":"LIMA: Less is more for alignment","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Zhou"},{"key":"ref15","first-page":"54111","article-title":"On evaluating adversarial robustness of large vision-language models","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Zhao"},{"key":"ref16","article-title":"Vision-LLMs can fool themselves with self-generated typographic attacks","author":"Qraitem","year":"2024","journal-title":"arXiv:2402.00626"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v39i22.34568"},{"key":"ref18","article-title":"Query-relevant images jailbreak large multi-modal models","author":"Liu","year":"2023","journal-title":"arXiv:2311.17600"},{"key":"ref19","first-page":"24824","article-title":"Chain-of-thought prompting elicits reasoning in large language models","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Wei"},{"key":"ref20","article-title":"Universal and transferable adversarial attacks on aligned language models","author":"Zou","year":"2023","journal-title":"arXiv:2307.15043"},{"key":"ref21","volume-title":"Gemini","year":"2024"},{"key":"ref22","volume-title":"Chatglm","year":"2023"},{"key":"ref23","volume-title":"Qwen","author":"Cloud","year":"2023"},{"key":"ref24","volume-title":"Ernie Bot","author":"Baidu","year":"2023"},{"key":"ref25","article-title":"Jailbreak and guard aligned language models with only few in-context demonstrations","author":"Wei","year":"2023","journal-title":"arXiv:2310.06387"},{"key":"ref26","article-title":"Tree of attacks: Jailbreaking black-box LLMs automatically","author":"Mehrotra","year":"2023","journal-title":"arXiv:2312.02119"},{"key":"ref27","article-title":"Scalable and transferable black-box jailbreaks for language models via persona modulation","author":"Shah","year":"2023","journal-title":"arXiv:2311.03348"},{"key":"ref28","article-title":"AutoDAN: Generating stealthy jailbreak prompts on aligned large language models","author":"Liu","year":"2023","journal-title":"arXiv:2310.04451"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v38i19.30150"},{"key":"ref30","article-title":"Jailbreaking attack against multimodal large language model","author":"Niu","year":"2024","journal-title":"arXiv:2402.02309"},{"key":"ref31","article-title":"InstructTA: Instruction-tuned targeted attack for large vision-language models","author":"Wang","year":"2023","journal-title":"arXiv:2312.01886"},{"key":"ref32","article-title":"How robust is Google\u2019s bard to adversarial image attacks?","author":"Dong","year":"2023","journal-title":"arXiv:2309.11751"},{"key":"ref33","article-title":"Fool your (vision and) language model with embarrassingly simple permutations","author":"Zong","year":"2023","journal-title":"arXiv:2310.01651"},{"key":"ref34","article-title":"MiniGPT-4: Enhancing vision-language understanding with advanced large language models","author":"Zhu","year":"2023","journal-title":"arXiv:2304.10592"},{"key":"ref35","first-page":"8748","article-title":"Learning transferable visual models from natural language supervision","volume-title":"Proc. 38th Int. Conf. Mach. Learn.","author":"Radford"},{"key":"ref36","article-title":"LLaMA: Open and efficient foundation language models","author":"Touvron","year":"2023","journal-title":"arXiv:2302.13971"},{"key":"ref37","volume-title":"Vicuna: An Open-Source Chatbot Impressing Gpt-4 With 90% Chatgpt Quality","author":"Chiang","year":"2023"},{"key":"ref38","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.findings-acl.775"},{"key":"ref39","article-title":"DRESS: Instructing large vision-language models to align and interact with humans via natural language feedback","author":"Chen","year":"2023","journal-title":"arXiv:2311.10081"},{"key":"ref40","article-title":"WARM: On the benefits of weight averaged reward models","author":"Ram\u00e9","year":"2024","journal-title":"arXiv:2401.12187"},{"key":"ref41","article-title":"Towards deep learning models resistant to adversarial attacks","author":"Madry","year":"2017","journal-title":"arXiv:1706.06083"},{"key":"ref42","article-title":"Navigate through enigmatic labyrinth a survey of chain of thought reasoning: Advances, frontiers and future","author":"Chu","year":"2023","journal-title":"arXiv:2309.15402"},{"key":"ref43","first-page":"22199","article-title":"Large language models are zero-shot reasoners","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Kojima"},{"key":"ref44","first-page":"49250","article-title":"InstructBLIP: Towards general-purpose vision-language models with instruction tuning","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Dai"},{"key":"ref45","article-title":"GPT-4 technical report","volume-title":"arXiv:2303.08774","author":"OpenAI","year":"2023"},{"key":"ref46","article-title":"Llama guard: LLM-based input\u2013output safeguard for human-AI conversations","author":"Inan","year":"2023","journal-title":"arXiv:2312.06674"},{"key":"ref47","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.emnlp-main.895"},{"key":"ref48","doi-asserted-by":"publisher","DOI":"10.1145\/3748239.3748242"},{"key":"ref49","article-title":"ImgTrojan: Jailbreaking vision-language models with ONE image","author":"Tao","year":"2024","journal-title":"arXiv:2403.02910"},{"key":"ref50","doi-asserted-by":"publisher","DOI":"10.1109\/TPAMI.2023.3347693"},{"key":"ref51","doi-asserted-by":"publisher","DOI":"10.1038\/s42256-023-00765-8"},{"key":"ref52","article-title":"SmoothLLM: Defending large language models against jailbreaking attacks","author":"Robey","year":"2023","journal-title":"arXiv:2310.03684"},{"key":"ref53","article-title":"JailGuard: A universal detection framework for LLM prompt-based attacks","author":"Zhang","year":"2023","journal-title":"arXiv:2312.10766"},{"key":"ref54","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-72643-9_23"},{"key":"ref55","article-title":"ETA: Evaluating then aligning safety of vision language models at inference time","author":"Ding","year":"2024","journal-title":"arXiv:2410.06625"},{"key":"ref56","article-title":"DeepInception: Hypnotize large language model to be jailbreaker","author":"Li","year":"2023","journal-title":"arXiv:2311.03191"},{"key":"ref57","first-page":"1","article-title":"Language models are few-shot learners","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Brown"},{"key":"ref58","doi-asserted-by":"publisher","DOI":"10.1109\/TIM.2003.815989"},{"key":"ref59","doi-asserted-by":"publisher","DOI":"10.1109\/TIP.2005.852196"},{"key":"ref60","doi-asserted-by":"publisher","DOI":"10.3115\/v1\/W14-1207"},{"key":"ref61","doi-asserted-by":"publisher","DOI":"10.1109\/ICICTA.2011.64"}],"container-title":["IEEE Transactions on Information Forensics and Security"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/10206\/10810755\/11059299.pdf?arnumber=11059299","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,7,18]],"date-time":"2025-07-18T04:40:40Z","timestamp":1752813640000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11059299\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025]]},"references-count":61,"URL":"https:\/\/doi.org\/10.1109\/tifs.2025.3583249","relation":{},"ISSN":["1556-6013","1556-6021"],"issn-type":[{"value":"1556-6013","type":"print"},{"value":"1556-6021","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025]]}}}