{"id":"https://openalex.org/W7140143686","doi":"https://doi.org/10.18653/v1/2026.findings-eacl.145","title":"Ensemble Privacy Defense for Knowledge-Intensive LLMs against Membership Inference Attacks","display_name":"Ensemble Privacy Defense for Knowledge-Intensive LLMs against Membership Inference Attacks","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7140143686","doi":"https://doi.org/10.18653/v1/2026.findings-eacl.145"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2026.findings-eacl.145","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-eacl.145","pdf_url":"https://aclanthology.org/2026.findings-eacl.145.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: EACL 2026","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2026.findings-eacl.145.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5130395694","display_name":"Haowei Fu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Haowei Fu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100541833","display_name":"Bo Ni","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bo Ni","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5130342338","display_name":"Han Xu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Han Xu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5130375009","display_name":"Kunpeng Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kunpeng Liu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5123396944","display_name":"Dan Lin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dan Lin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5036086705","display_name":"Tyler Derr","orcid":"https://orcid.org/0000-0002-0080-5998"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tyler Derr","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"2786","last_page":"2799"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10764","display_name":"Privacy-Preserving Technologies in Data","score":0.41839998960494995,"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/T10764","display_name":"Privacy-Preserving Technologies in Data","score":0.41839998960494995,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.12999999523162842,"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/T10237","display_name":"Cryptography and Data Security","score":0.07039999961853027,"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/inference","display_name":"Inference","score":0.4440000057220459},{"id":"https://openalex.org/keywords/information-privacy","display_name":"Information privacy","score":0.3239000141620636},{"id":"https://openalex.org/keywords/privacy-protection","display_name":"Privacy protection","score":0.30320000648498535},{"id":"https://openalex.org/keywords/confidentiality","display_name":"Confidentiality","score":0.3025999963283539},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.29910001158714294}],"concepts":[{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.5990999937057495},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5392000079154968},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.4440000057220459},{"id":"https://openalex.org/C108827166","wikidata":"https://www.wikidata.org/wiki/Q175975","display_name":"Internet privacy","level":1,"score":0.4081999957561493},{"id":"https://openalex.org/C123201435","wikidata":"https://www.wikidata.org/wiki/Q456632","display_name":"Information privacy","level":2,"score":0.3239000141620636},{"id":"https://openalex.org/C3017597292","wikidata":"https://www.wikidata.org/wiki/Q25052250","display_name":"Privacy protection","level":2,"score":0.30320000648498535},{"id":"https://openalex.org/C71745522","wikidata":"https://www.wikidata.org/wiki/Q2476929","display_name":"Confidentiality","level":2,"score":0.3025999963283539},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.29910001158714294},{"id":"https://openalex.org/C144133560","wikidata":"https://www.wikidata.org/wiki/Q4830453","display_name":"Business","level":0,"score":0.2777999937534332},{"id":"https://openalex.org/C95713431","wikidata":"https://www.wikidata.org/wiki/Q631425","display_name":"Vulnerability (computing)","level":2,"score":0.26170000433921814},{"id":"https://openalex.org/C12174686","wikidata":"https://www.wikidata.org/wiki/Q1058438","display_name":"Risk assessment","level":2,"score":0.25679999589920044}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2026.findings-eacl.145","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-eacl.145","pdf_url":"https://aclanthology.org/2026.findings-eacl.145.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: EACL 2026","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.findings-eacl.145","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-eacl.145","pdf_url":"https://aclanthology.org/2026.findings-eacl.145.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: EACL 2026","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"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/W7140143686.pdf","grobid_xml":"https://content.openalex.org/works/W7140143686.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Retrieval-Augmented":[0],"Generation":[1],"(RAG)":[2],"and":[3,61,100,111,136],"Supervised":[4],"Finetuning":[5],"(SFT)":[6],"have":[7],"become":[8],"the":[9,72,83,102],"predominant":[10],"paradigms":[11],"for":[12,21,134,138],"equipping":[13],"Large":[14],"Language":[15],"Models":[16],"(LLMs)":[17],"with":[18],"external":[19],"knowledge":[20,27],"diverse,":[22],"knowledge-intensive":[23],"tasks.However,":[24],"while":[25,144],"such":[26],"injection":[28],"improves":[29],"performance,":[30],"it":[31],"also":[32],"exposes":[33],"new":[34],"attack":[35],"surfaces.Membership":[36],"Inference":[37],"Attacks":[38],"(MIAs),":[39],"which":[40,98],"aim":[41],"to":[42,59,78,81,116,132,141],"determine":[43],"whether":[44],"a":[45,52,89,105,108,112],"given":[46],"data":[47],"sample":[48],"was":[49],"included":[50],"in":[51,63],"model's":[53],"training":[54],"set,":[55],"pose":[56],"serious":[57],"threats":[58],"privacy":[60,84],"trust":[62],"sensitive":[64],"domains.To":[65],"this":[66],"end,":[67],"we":[68,86],"first":[69],"systematically":[70],"evaluate":[71],"vulnerability":[73],"of":[74,104],"RAG-and":[75],"SFT-based":[76],"LLMs":[77],"various":[79],"MIAs.Then,":[80],"address":[82],"risk,":[85],"further":[87],"introduce":[88],"novel,":[90],"model-agnostic":[91],"defense":[92],"framework,":[93],"Ensemble":[94],"Privacy":[95],"Defense":[96],"(EPD),":[97],"aggregates":[99],"evaluates":[101],"outputs":[103],"knowledge-injected":[106],"LLM,":[107,110],"base":[109],"dedicated":[113],"judge":[114],"model":[115],"enhance":[117],"resistance":[118],"against":[119],"MIAs.Comprehensive":[120],"experiments":[121],"show":[122],"that,":[123],"on":[124],"average,":[125],"EPD":[126],"reduces":[127],"MIA":[128],"success":[129],"by":[130],"up":[131],"27.8%":[133],"SFT":[135],"526.3%":[137],"RAG":[139],"compared":[140],"inference-time":[142],"baseline,":[143],"maintaining":[145],"answer":[146],"quality.":[147]},"counts_by_year":[],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2026-03-24T00:00:00"}