{"id":"https://openalex.org/W7138124342","doi":"https://doi.org/10.1609/aaai.v40i42.40887","title":"Dynamic Deep Prompt Optimization for Defending Against Jailbreak Attacks on LLMs","display_name":"Dynamic Deep Prompt Optimization for Defending Against Jailbreak Attacks on LLMs","publication_year":2026,"publication_date":"2026-03-14","ids":{"openalex":"https://openalex.org/W7138124342","doi":"https://doi.org/10.1609/aaai.v40i42.40887"},"language":null,"primary_location":{"id":"doi:10.1609/aaai.v40i42.40887","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i42.40887","pdf_url":"https://ojs.aaai.org/index.php/AAAI/article/download/40887/44848","source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"diamond","oa_url":"https://ojs.aaai.org/index.php/AAAI/article/download/40887/44848","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5125754449","display_name":"Doniyorkhon Obidov","orcid":null},"institutions":[{"id":"https://openalex.org/I11957088","display_name":"Michigan Technological University","ror":"https://ror.org/0036rpn28","country_code":"US","type":"education","lineage":["https://openalex.org/I11957088"]}],"countries":["US"],"is_corresponding":true,"raw_author_name":"Doniyorkhon Obidov","raw_affiliation_strings":["Michigan Technological University"],"affiliations":[{"raw_affiliation_string":"Michigan Technological University","institution_ids":["https://openalex.org/I11957088"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5128000734","display_name":"Honggang Yu","orcid":null},"institutions":[{"id":"https://openalex.org/I83328450","display_name":"Miami University","ror":"https://ror.org/05nbqxr67","country_code":"US","type":"education","lineage":["https://openalex.org/I83328450"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Honggang Yu","raw_affiliation_strings":["Miami University"],"affiliations":[{"raw_affiliation_string":"Miami University","institution_ids":["https://openalex.org/I83328450"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129749439","display_name":"Xiaolong Guo","orcid":null},"institutions":[{"id":"https://openalex.org/I189590672","display_name":"Kansas State University","ror":"https://ror.org/05p1j8758","country_code":"US","type":"education","lineage":["https://openalex.org/I189590672"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Xiaolong Guo","raw_affiliation_strings":["Kansas State University"],"affiliations":[{"raw_affiliation_string":"Kansas State University","institution_ids":["https://openalex.org/I189590672"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5011038036","display_name":"Kaichen Yang","orcid":"https://orcid.org/0000-0003-1027-6708"},"institutions":[{"id":"https://openalex.org/I11957088","display_name":"Michigan Technological University","ror":"https://ror.org/0036rpn28","country_code":"US","type":"education","lineage":["https://openalex.org/I11957088"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Kaichen Yang","raw_affiliation_strings":["Michigan Technological University"],"affiliations":[{"raw_affiliation_string":"Michigan Technological University","institution_ids":["https://openalex.org/I11957088"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":4,"corresponding_author_ids":["https://openalex.org/A5125754449"],"corresponding_institution_ids":["https://openalex.org/I11957088"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.34256793,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"40","issue":"42","first_page":"35742","last_page":"35750"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.899399995803833,"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":0.899399995803833,"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/T11241","display_name":"Advanced Malware Detection Techniques","score":0.039000000804662704,"subfield":{"id":"https://openalex.org/subfields/1711","display_name":"Signal Processing"},"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":0.010900000110268593,"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/set","display_name":"Set (abstract data type)","score":0.6622999906539917},{"id":"https://openalex.org/keywords/adaptability","display_name":"Adaptability","score":0.5860000252723694},{"id":"https://openalex.org/keywords/feature","display_name":"Feature (linguistics)","score":0.41200000047683716},{"id":"https://openalex.org/keywords/unintended-consequences","display_name":"Unintended consequences","score":0.398499995470047},{"id":"https://openalex.org/keywords/prefix","display_name":"Prefix","score":0.3619000017642975},{"id":"https://openalex.org/keywords/optimization-problem","display_name":"Optimization problem","score":0.30979999899864197}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7206000089645386},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.6622999906539917},{"id":"https://openalex.org/C177606310","wikidata":"https://www.wikidata.org/wiki/Q5674297","display_name":"Adaptability","level":2,"score":0.5860000252723694},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.47099998593330383},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.41200000047683716},{"id":"https://openalex.org/C2776889888","wikidata":"https://www.wikidata.org/wiki/Q1135789","display_name":"Unintended consequences","level":2,"score":0.398499995470047},{"id":"https://openalex.org/C112930515","wikidata":"https://www.wikidata.org/wiki/Q4389547","display_name":"Risk analysis (engineering)","level":1,"score":0.3813000023365021},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.37950000166893005},{"id":"https://openalex.org/C141603448","wikidata":"https://www.wikidata.org/wiki/Q134830","display_name":"Prefix","level":2,"score":0.3619000017642975},{"id":"https://openalex.org/C137836250","wikidata":"https://www.wikidata.org/wiki/Q984063","display_name":"Optimization problem","level":2,"score":0.30979999899864197},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.3084000051021576},{"id":"https://openalex.org/C140547941","wikidata":"https://www.wikidata.org/wiki/Q7797194","display_name":"Threat model","level":2,"score":0.28760001063346863},{"id":"https://openalex.org/C2777093003","wikidata":"https://www.wikidata.org/wiki/Q6508345","display_name":"Lead (geology)","level":2,"score":0.28040000796318054},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.2743000090122223},{"id":"https://openalex.org/C33009525","wikidata":"https://www.wikidata.org/wiki/Q208841","display_name":"Coevolution","level":2,"score":0.2709999978542328},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.26759999990463257}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1609/aaai.v40i42.40887","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i42.40887","pdf_url":"https://ojs.aaai.org/index.php/AAAI/article/download/40887/44848","source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"}],"best_oa_location":{"id":"doi:10.1609/aaai.v40i42.40887","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i42.40887","pdf_url":"https://ojs.aaai.org/index.php/AAAI/article/download/40887/44848","source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/16","display_name":"Peace, Justice and strong institutions","score":0.7928236126899719}],"awards":[{"id":"https://openalex.org/G106460892","display_name":null,"funder_award_id":"2419880","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G1743096580","display_name":"ERI: Towards Robust and Secure Intelligent 3D Sensing Systems","funder_award_id":"2347426","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G848032724","display_name":null,"funder_award_id":"Science","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"}],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7138124342.pdf","grobid_xml":"https://content.openalex.org/works/W7138124342.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"Language":[1],"Models":[2],"(LLMs)":[3],"demonstrate":[4,144],"impressive":[5],"capabilities":[6],"across":[7],"many":[8],"applications":[9],"but":[10],"remain":[11],"vulnerable":[12],"to":[13,36,64,96],"jailbreak":[14,78],"attacks,":[15],"which":[16],"elicit":[17],"harmful":[18,170],"or":[19,59],"unintended":[20],"content.":[21],"While":[22],"model":[23],"fine-tuning":[24],"is":[25,32],"an":[26,118],"option":[27],"for":[28],"safety":[29],"alignment,":[30],"it":[31],"costly":[33],"and":[34,66,142,158],"prone":[35],"catastrophic":[37],"forgetting.":[38],"Prompt":[39,73],"optimization":[40,151],"has":[41],"emerged":[42],"as":[43,93],"a":[44,102,113,137],"promising":[45],"alternative,":[46],"yet":[47],"existing":[48],"prompt-based":[49],"defenses":[50],"typically":[51],"rely":[52],"on":[53,81,136,154],"static":[54,149],"modifications":[55],"(e.g.,":[56],"fixed":[57],"prefixes":[58],"suffixes)":[60],"that":[61,145],"cannot":[62],"adapt":[63],"diverse":[65,138],"evolving":[67],"attacks.":[68],"We":[69],"propose":[70],"Dynamic":[71],"Deep":[72],"Optimization":[74],"(DDPO),":[75],"the":[76,87,123],"first":[77],"defense":[79,120],"based":[80],"deep":[82],"prompt":[83,150],"optimization.":[84],"DDPO":[85,146],"uses":[86],"target":[88],"LLM's":[89,124],"own":[90],"intermediate":[91,115],"layers":[92],"feature":[94],"extractors":[95],"dynamically":[97],"generate":[98],"defensive":[99],"embeddings":[100,108],"via":[101],"lightweight":[103],"multilayer":[104],"perceptron.":[105],"These":[106],"tailored":[107],"are":[109],"then":[110],"injected":[111],"into":[112],"subsequent":[114],"layer,":[116],"enabling":[117],"input-dependent":[119],"without":[121],"modifying":[122],"weights.":[125],"This":[126],"design":[127],"ensures":[128],"high":[129],"adaptability":[130],"with":[131],"minimal":[132],"computational":[133],"overhead.":[134],"Experiments":[135],"set":[139],"of":[140],"models":[141,157],"attacks":[143],"significantly":[147],"outperforms":[148],"methods,":[152],"particularly":[153],"weakly":[155],"aligned":[156],"when":[159],"handling":[160],"semantically":[161],"ambiguous":[162],"benign":[163],"prompts,":[164],"successfully":[165],"distinguishing":[166],"them":[167],"from":[168],"genuinely":[169],"requests.":[171]},"counts_by_year":[],"updated_date":"2026-04-03T22:45:19.894376","created_date":"2026-03-18T00:00:00"}