{"id":"https://openalex.org/W4404782637","doi":"https://doi.org/10.18653/v1/2024.emnlp-main.1014","title":"User Inference Attacks on Large Language Models","display_name":"User Inference Attacks on Large Language Models","publication_year":2024,"publication_date":"2024-01-01","ids":{"openalex":"https://openalex.org/W4404782637","doi":"https://doi.org/10.18653/v1/2024.emnlp-main.1014"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2024.emnlp-main.1014","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2024.emnlp-main.1014","pdf_url":"https://aclanthology.org/2024.emnlp-main.1014.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":"Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2024.emnlp-main.1014.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5056645742","display_name":"Nikhil Kandpal","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Nikhil Kandpal","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5016535318","display_name":"Krishna Pillutla","orcid":"https://orcid.org/0000-0002-1262-8466"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Krishna Pillutla","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5111219674","display_name":"Alina Oprea","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Alina Oprea","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5064699160","display_name":"Peter Kairouz","orcid":"https://orcid.org/0000-0001-6897-5937"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Peter Kairouz","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5063017186","display_name":"Christopher A. Choquette-Choo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Christopher A. Choquette-Choo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5100413782","display_name":"Zheng Xu","orcid":"https://orcid.org/0000-0003-0648-0327"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zheng Xu","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":16,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"18238","last_page":"18265"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.9685999751091003,"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/T10028","display_name":"Topic Modeling","score":0.9685999751091003,"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/T13702","display_name":"Machine Learning in Healthcare","score":0.9659000039100647,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.9132999777793884,"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/computer-science","display_name":"Computer science","score":0.8047908544540405},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.66322261095047},{"id":"https://openalex.org/keywords/natural-language-processing","display_name":"Natural language processing","score":0.4563830494880676},{"id":"https://openalex.org/keywords/programming-language","display_name":"Programming language","score":0.4540501534938812},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.36097925901412964}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8047908544540405},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.66322261095047},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.4563830494880676},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.4540501534938812},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.36097925901412964}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2024.emnlp-main.1014","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2024.emnlp-main.1014","pdf_url":"https://aclanthology.org/2024.emnlp-main.1014.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":"Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2024.emnlp-main.1014","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2024.emnlp-main.1014","pdf_url":"https://aclanthology.org/2024.emnlp-main.1014.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":"Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4404782637.pdf","grobid_xml":"https://content.openalex.org/works/W4404782637.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W2390279801","https://openalex.org/W4391913857","https://openalex.org/W2358668433","https://openalex.org/W4396701345","https://openalex.org/W2376932109","https://openalex.org/W2001405890","https://openalex.org/W3204019825"],"abstract_inverted_index":{"Text":[0],"written":[1],"by":[2,215],"humans":[3],"makes":[4],"up":[5],"the":[6,10,86,115,121,183,193],"vast":[7],"majority":[8],"of":[9,22,52,68,154],"data":[11,55,80,210],"used":[12,57,82],"to":[13,34,48,58,76,101,107,139,182,195,198],"pre-train":[14],"and":[15,30,110,130,149,177],"finetune":[16],"large":[17],"language":[18],"models":[19],"(LLMs).Many":[20],"sources":[21],"this":[23,40,64],"data-like":[24],"code,":[25],"forum":[26],"posts,":[27],"personal":[28],"websites,":[29],"books-are":[31],"easily":[32],"attributed":[33],"one":[35],"or":[36],"a":[37,53,66,111,151,179],"few":[38,112],"\"users\".In":[39],"paper,":[41],"we":[42,159],"ask":[43],"if":[44,50,258],"it":[45,71],"is":[46],"possible":[47],"infer":[49],"any":[51],"user's":[54,180],"was":[56,81,255],"train":[59],"an":[60,108,143],"LLM.Not":[61],"only":[62,104],"would":[63,72],"constitute":[65],"breach":[67],"privacy,":[69,172],"but":[70],"also":[73],"enable":[74],"users":[75,136],"detect":[77],"when":[78],"their":[79],"for":[83,90,163],"training.We":[84],"develop":[85,196],"first":[87],"effective":[88],"attacks":[89,98],"user":[91,140,165,203],"inferenceat":[92],"times,":[93],"with":[94,169],"near-perfect":[95],"success-against":[96],"LLMs.Our":[97],"are":[99],"easy":[100],"employ,":[102],"requiring":[103],"black-box":[105],"access":[106,218],"LLM":[109,205,207],"samples":[113,212],"from":[114,202,271],"user,":[116],"which":[117],"need":[118,194],"not":[119],"be":[120],"ones":[122],"that":[123,132],"were":[124],"trained":[125],"on.We":[126],"find,":[127],"both":[128],"theoretically":[129],"empirically,":[131],"certain":[133],"properties":[134],"make":[135],"more":[137],"susceptible":[138],"inference:":[141],"being":[142],"outlier,":[144],"having":[145],"highly":[146],"correlated":[147],"examples,":[148,176],"contributing":[150],"larger":[152],"fraction":[153],"data.Based":[155],"on":[156],"these":[157,186],"findings,":[158],"identify":[160],"several":[161],"methods":[162,197],"mitigating":[164],"inference":[166],"including":[167],"training":[168,184,257],"example-level":[170],"differential":[171],"removing":[173],"within-user":[174],"duplicate":[175],"reducing":[178],"contribution":[181],"data.Though":[185],"provide":[187],"partial":[188],"mitigation,":[189],"our":[190],"work":[191],"highlights":[192],"fully":[199],"protect":[200],"LLMs":[201],"inference.Pre-trained":[204],"Finetuned":[206],"!User-level":[208],"finetuned":[209],"Training":[211],"Samples":[213],"known":[214],"attacker":[216],"Query":[217],"Adversary":[219],"Target":[220],"User":[221],"2.":[222],"For":[223],"each":[224],"(#)":[225,228,247,252],"compute":[226],"!(":[227],")":[229,248,253],"3.":[230],"Test":[231],"statistic":[232],"4":[233,259],"(%)":[234,260,267],",":[235,236,261,262,268,269],"(&)":[237,263],"=":[238],"%":[239],"&":[240,242,270],"#'%":[241],"log":[243],"(":[244,249],"!":[245],"()":[246,251],"%&'":[250],"4.":[254],"in":[256],">":[264],"1.":[265],"Sample":[266],"+":[272]},"counts_by_year":[{"year":2026,"cited_by_count":4},{"year":2025,"cited_by_count":8},{"year":2024,"cited_by_count":3},{"year":2023,"cited_by_count":1}],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}