{"id":"https://openalex.org/W4412888545","doi":"https://doi.org/10.18653/v1/2025.findings-acl.346","title":"CAVGAN: Unifying Jailbreak and Defense of LLMs via Generative Adversarial Attacks on their Internal Representations","display_name":"CAVGAN: Unifying Jailbreak and Defense of LLMs via Generative Adversarial Attacks on their Internal Representations","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412888545","doi":"https://doi.org/10.18653/v1/2025.findings-acl.346"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.findings-acl.346","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.346","pdf_url":"https://aclanthology.org/2025.findings-acl.346.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: ACL 2025","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.findings-acl.346.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100631865","display_name":"Xiaohu Li","orcid":"https://orcid.org/0000-0001-8278-3878"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiaohu Li","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5119181444","display_name":"Yunfeng Ning","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yunfeng Ning","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5119181445","display_name":"Zepeng Bao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zepeng Bao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5022569658","display_name":"Mayi Xu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mayi Xu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5006836653","display_name":"Jianhao Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jianhao Chen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5040759280","display_name":"Tieyun Qian","orcid":"https://orcid.org/0000-0003-4667-5794"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tieyun Qian","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":6,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":3.3614,"has_fulltext":true,"cited_by_count":2,"citation_normalized_percentile":{"value":0.92759087,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":91,"max":98},"biblio":{"volume":null,"issue":null,"first_page":"6664","last_page":"6678"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.9983999729156494,"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":0.9983999729156494,"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/T12034","display_name":"Digital and Cyber Forensics","score":0.934499979019165,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/adversarial-system","display_name":"Adversarial system","score":0.8584514856338501},{"id":"https://openalex.org/keywords/generative-grammar","display_name":"Generative grammar","score":0.6352595686912537},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.5106542706489563},{"id":"https://openalex.org/keywords/generative-adversarial-network","display_name":"Generative adversarial network","score":0.44861969351768494},{"id":"https://openalex.org/keywords/computer-security","display_name":"Computer security","score":0.43829891085624695},{"id":"https://openalex.org/keywords/cognitive-science","display_name":"Cognitive science","score":0.39133188128471375},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.26964646577835083},{"id":"https://openalex.org/keywords/psychology","display_name":"Psychology","score":0.19508305191993713}],"concepts":[{"id":"https://openalex.org/C37736160","wikidata":"https://www.wikidata.org/wiki/Q1801315","display_name":"Adversarial system","level":2,"score":0.8584514856338501},{"id":"https://openalex.org/C39890363","wikidata":"https://www.wikidata.org/wiki/Q36108","display_name":"Generative grammar","level":2,"score":0.6352595686912537},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5106542706489563},{"id":"https://openalex.org/C2988773926","wikidata":"https://www.wikidata.org/wiki/Q25104379","display_name":"Generative adversarial network","level":3,"score":0.44861969351768494},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.43829891085624695},{"id":"https://openalex.org/C188147891","wikidata":"https://www.wikidata.org/wiki/Q147638","display_name":"Cognitive science","level":1,"score":0.39133188128471375},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.26964646577835083},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.19508305191993713},{"id":"https://openalex.org/C115961682","wikidata":"https://www.wikidata.org/wiki/Q860623","display_name":"Image (mathematics)","level":2,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.findings-acl.346","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.346","pdf_url":"https://aclanthology.org/2025.findings-acl.346.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: ACL 2025","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.findings-acl.346","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.346","pdf_url":"https://aclanthology.org/2025.findings-acl.346.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: ACL 2025","raw_type":"proceedings-article"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/16","display_name":"Peace, Justice and strong institutions","score":0.46000000834465027}],"awards":[{"id":"https://openalex.org/G6102162879","display_name":null,"funder_award_id":"2042022dx0001","funder_id":"https://openalex.org/F4320335787","funder_display_name":"Fundamental Research Funds for the Central Universities"},{"id":"https://openalex.org/G672599822","display_name":null,"funder_award_id":"62276193","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G8830422215","display_name":null,"funder_award_id":"2042022dx0001","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"}],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"},{"id":"https://openalex.org/F4320335787","display_name":"Fundamental Research Funds for the Central Universities","ror":null}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412888545.pdf","grobid_xml":"https://content.openalex.org/works/W4412888545.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W2502115930","https://openalex.org/W2482350142","https://openalex.org/W4246396837","https://openalex.org/W2888032422","https://openalex.org/W2996316059","https://openalex.org/W4410783345","https://openalex.org/W4377980832","https://openalex.org/W2897769091","https://openalex.org/W2845413374","https://openalex.org/W3005996785"],"abstract_inverted_index":{"Security":[0],"alignment":[1],"enables":[2],"the":[3,10,21,36,41,56,68,83,93,98,126,131,143,153],"Large":[4],"Language":[5],"Model":[6],"(LLM)":[7],"to":[8,75,82,91,100],"gain":[9],"protection":[11,38],"against":[12],"malicious":[13],"queries,":[14],"but":[15,148],"various":[16],"jailbreak":[17,31,71,103,116,133],"attack":[18,49,104],"methods":[19],"reveal":[20],"vulnerability":[22],"of":[23,40,60,70,119,138,145,157],"this":[24],"security":[25,37,94,155,165],"mechanism.Previous":[26],"studies":[27],"have":[28],"isolated":[29],"LLM":[30,61,99],"attacks":[32],"and":[33,43,50,79,105],"defenses.We":[34],"analyze":[35],"mechanism":[39],"LLM,":[42],"propose":[44],"a":[45],"framework":[46],"that":[47,110],"combines":[48],"defense.Our":[51],"method":[52,112],"is":[53],"based":[54],"on":[55,130,152],"linearly":[57],"separable":[58],"property":[59],"intermediate":[62],"layer":[63],"embedding,":[64],"as":[65,67],"well":[66],"essence":[69],"attack,":[72],"which":[73],"aims":[74],"embed":[76],"harmful":[77,172],"problems":[78],"transfer":[80],"them":[81],"safe":[84],"area.We":[85],"utilize":[86],"generative":[87],"adversarial":[88],"network":[89],"(GAN)":[90],"learn":[92],"judgment":[95],"boundary":[96],"inside":[97],"achieve":[101],"efficient":[102],"defense.The":[106],"experimental":[107],"results":[108],"indicate":[109],"our":[111,146],"achieves":[113],"an":[114,136],"average":[115,137],"success":[117,128],"rate":[118,129],"88.85%":[120],"across":[121],"three":[122],"popular":[123],"LLMs,":[124,158],"while":[125],"defense":[127],"state-ofthe-art":[132],"dataset":[134],"reaches":[135],"84.17%.This":[139],"not":[140],"only":[141],"validates":[142],"effectiveness":[144],"approach":[147],"also":[149],"sheds":[150],"light":[151],"internal":[154],"mechanisms":[156],"offering":[159],"new":[160],"insights":[161],"for":[162],"enhancing":[163],"model":[164],"1":[166],".Warning:":[167],"This":[168],"paper":[169],"contains":[170],"some":[171],"text":[173],"examples.":[174]},"counts_by_year":[{"year":2026,"cited_by_count":1},{"year":2025,"cited_by_count":1}],"updated_date":"2026-06-21T07:57:09.225873","created_date":"2025-10-10T00:00:00"}