{"id":"https://openalex.org/W4412945189","doi":"https://doi.org/10.18653/v1/2025.acl-long.724","title":"HiddenDetect: Detecting Jailbreak Attacks against Multimodal Large Language Models via Monitoring Hidden States","display_name":"HiddenDetect: Detecting Jailbreak Attacks against Multimodal Large Language Models via Monitoring Hidden States","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412945189","doi":"https://doi.org/10.18653/v1/2025.acl-long.724"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.acl-long.724","doi":"https://doi.org/10.18653/v1/2025.acl-long.724","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.724","pdf_url":"https://aclanthology.org/2025.acl-long.724.pdf","source":{"id":"https://openalex.org/S4393915740","display_name":"Proceedings of the conference - Association for Computational Linguistics. Meeting","issn_l":"0736-587X","issn":["0736-587X"],"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":["norway-1"],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"volume":null,"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"bronze","oa_url":"https://aclanthology.org/2025.acl-long.724.pdf","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5102370506","display_name":"Yilei Jiang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I177725633","display_name":"Chinese University of Hong Kong","ror":"https://ror.org/00t33hh48","country_code":"HK","type":"education","lineage":["https://openalex.org/I177725633"]},{"id":"https://openalex.org/I45928872","display_name":"Alibaba Group (China)","ror":"https://ror.org/00k642b80","country_code":"CN","type":"company","lineage":["https://openalex.org/I45928872"]}],"countries":["CN","HK"],"is_corresponding":false,"raw_author_name":"Yilei Jiang","raw_affiliation_strings":["Future Lab , Alibaba Group","MMLab , The Chinese University of Hong Kong"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Future Lab , Alibaba Group","institution_ids":["https://openalex.org/I45928872"]},{"raw_affiliation_string":"MMLab , The Chinese University of Hong Kong","institution_ids":["https://openalex.org/I177725633"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Xinyan Gao","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I177725633","display_name":"Chinese University of Hong Kong","ror":"https://ror.org/00t33hh48","country_code":"HK","type":"education","lineage":["https://openalex.org/I177725633"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Xinyan Gao","raw_affiliation_strings":["MMLab , The Chinese University of Hong Kong"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"MMLab , The Chinese University of Hong Kong","institution_ids":["https://openalex.org/I177725633"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5114139333","display_name":"Tianshuo Peng","orcid":"https://orcid.org/0009-0009-7347-0648","observed_orcids":["https://orcid.org/0009-0009-7347-0648"]},"institutions":[{"id":"https://openalex.org/I177725633","display_name":"Chinese University of Hong Kong","ror":"https://ror.org/00t33hh48","country_code":"HK","type":"education","lineage":["https://openalex.org/I177725633"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Tianshuo Peng","raw_affiliation_strings":["MMLab , The Chinese University of Hong Kong"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"MMLab , The Chinese University of Hong Kong","institution_ids":["https://openalex.org/I177725633"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5066811955","display_name":"Y. H. Tan","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I45928872","display_name":"Alibaba Group (China)","ror":"https://ror.org/00k642b80","country_code":"CN","type":"company","lineage":["https://openalex.org/I45928872"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yingshui Tan","raw_affiliation_strings":["Future Lab , Alibaba Group"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Future Lab , Alibaba Group","institution_ids":["https://openalex.org/I45928872"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5047643306","display_name":"Xiaoyong Zhu","orcid":"https://orcid.org/0000-0001-6333-3329","observed_orcids":["https://orcid.org/0000-0001-6333-3329"]},"institutions":[{"id":"https://openalex.org/I45928872","display_name":"Alibaba Group (China)","ror":"https://ror.org/00k642b80","country_code":"CN","type":"company","lineage":["https://openalex.org/I45928872"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xiaoyong Zhu","raw_affiliation_strings":["Future Lab , Alibaba Group"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Future Lab , Alibaba Group","institution_ids":["https://openalex.org/I45928872"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Bo Zheng","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I45928872","display_name":"Alibaba Group (China)","ror":"https://ror.org/00k642b80","country_code":"CN","type":"company","lineage":["https://openalex.org/I45928872"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Bo Zheng","raw_affiliation_strings":["Future Lab , Alibaba Group"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Future Lab , Alibaba Group","institution_ids":["https://openalex.org/I45928872"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5078165161","display_name":"Xiangyu Yue","orcid":"https://orcid.org/0000-0002-6887-2046","observed_orcids":["https://orcid.org/0000-0002-6887-2046"]},"institutions":[{"id":"https://openalex.org/I177725633","display_name":"Chinese University of Hong Kong","ror":"https://ror.org/00t33hh48","country_code":"HK","type":"education","lineage":["https://openalex.org/I177725633"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Xiangyu Yue","raw_affiliation_strings":["MMLab , The Chinese University of Hong Kong"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"MMLab , The Chinese University of Hong Kong","institution_ids":["https://openalex.org/I177725633"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":12.1429,"has_fulltext":true,"cited_by_count":13,"citation_normalized_percentile":{"value":0.98999321,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":90,"max":100},"biblio":{"volume":null,"issue":null,"first_page":"14880","last_page":"14893"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.9926999807357788,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.9926999807357788,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11424","display_name":"Security and Verification in Computing","score":0.006200000178068876,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10764","display_name":"Privacy-Preserving Technologies in Data","score":9.999999747378752e-05,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/jailbreak-attacks","display_name":"jailbreak attacks","score":0.9710000157356262},{"id":"https://openalex.org/keywords/multimodal-large-language-models","display_name":"multimodal large language models","score":0.9819999933242798},{"id":"https://openalex.org/keywords/attack-detection","display_name":"attack detection","score":0.8330000042915344},{"id":"https://openalex.org/keywords/large-language-model-security","display_name":"large language model security","score":0.5609999895095825},{"id":"https://openalex.org/keywords/multimodal-models","display_name":"multimodal models","score":0.39100000262260437}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7415592074394226},{"id":"https://openalex.org/C23224414","wikidata":"https://www.wikidata.org/wiki/Q176769","display_name":"Hidden Markov model","level":2,"score":0.5131047964096069},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.35313600301742554},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.3408086597919464}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.acl-long.724","doi":"https://doi.org/10.18653/v1/2025.acl-long.724","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.724","pdf_url":"https://aclanthology.org/2025.acl-long.724.pdf","source":{"id":"https://openalex.org/S4393915740","display_name":"Proceedings of the conference - Association for Computational Linguistics. Meeting","issn_l":"0736-587X","issn":["0736-587X"],"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":["norway-1"],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.acl-long.724","doi":"https://doi.org/10.18653/v1/2025.acl-long.724","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.724","pdf_url":"https://aclanthology.org/2025.acl-long.724.pdf","source":{"id":"https://openalex.org/S4393915740","display_name":"Proceedings of the conference - Association for Computational Linguistics. Meeting","issn_l":"0736-587X","issn":["0736-587X"],"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":["norway-1"],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412945189.pdf","grobid_xml":"https://content.openalex.org/works/W4412945189.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7167168612","https://openalex.org/W7160067437","https://openalex.org/W7109764601","https://openalex.org/W4415341465","https://openalex.org/W7166859271","https://openalex.org/W7163867469","https://openalex.org/W7160089530","https://openalex.org/W7168187268","https://openalex.org/W4416034042","https://openalex.org/W4416215218"],"abstract_inverted_index":{"The":[0],"integration":[1],"of":[2,8],"additional":[3],"modalities":[4],"increases":[5],"the":[6,33],"susceptibility":[7],"large":[9],"visionlanguage":[10],"models":[11],"(LVLMs)":[12],"to":[13,21,75,100],"safety":[14,35],"risks,":[15],"such":[16],"as":[17],"jailbreak":[18,112],"attacks,":[19],"compared":[20],"their":[22,54],"language-only":[23],"counterparts.While":[24],"existing":[25],"research":[26],"primarily":[27],"focuses":[28],"on":[29,85],"post-hoc":[30],"alignment":[31],"techniques,":[32],"underlying":[34],"mechanisms":[36],"within":[37,53],"LVLMs":[38,48,62],"remain":[39],"largely":[40],"unexplored.In":[41],"this":[42,86,143],"work":[43],",":[44],"we":[45,88],"investigate":[46],"whether":[47],"inherently":[49],"encode":[50],"safety-relevant":[51],"signals":[52],"internal":[55,97],"activations":[56,99],"during":[57],"inference.Our":[58],"findings":[59],"reveal":[60],"that":[61,95,105,148],"exhibit":[63],"distinct":[64],"activation":[65],"patterns":[66],"when":[67],"processing":[68],"unsafe":[69],"prompts,":[70],"which":[71],"can":[72],"be":[73,137,150],"leveraged":[74],"detect":[76],"and":[77,125],"mitigate":[78],"adversarial":[79],"inputs":[80],"without":[81],"requiring":[82],"extensive":[83],"fine-tuning.Building":[84],"insight,":[87],"introduce":[89],"HiddenDetect,":[90],"a":[91],"novel":[92],"tuning-free":[93],"framework":[94],"harnesses":[96],"model":[98],"enhance":[101],"safety.Experimental":[102],"results":[103],"show":[104],"HiddenDetect":[106],"surpasses":[107],"state-of-the-art":[108],"methods":[109],"in":[110],"detecting":[111],"attacks":[113],"against":[114,132],"LVLMs.By":[115],"utilizing":[116],"intrinsic":[117],"safety-aware":[118],"patterns,":[119],"our":[120],"method":[121],"provides":[122],"an":[123],"efficient":[124],"scalable":[126],"solution":[127],"for":[128],"strengthening":[129],"LVLM":[130],"robustness":[131],"multimodal":[133],"threats.Our":[134],"code":[135],"will":[136],"released":[138],"publicly":[139],"at":[140],"https://github.com/leigest519/":[141],"HiddenDetect.Warning:":[142],"paper":[144],"contains":[145],"example":[146],"data":[147],"may":[149],"offensive":[151],"or":[152],"harmful.":[153]},"counts_by_year":[{"year":2026,"cited_by_count":12},{"year":2025,"cited_by_count":1}],"updated_date":"2026-11-03T10:50:19.383937","created_date":"2025-10-10T00:00:00"}