{"id":"https://openalex.org/W7204154663","doi":"https://doi.org/10.48550/arxiv.2608.21656","title":"Mitigating Database Leakage in RAG Systems with Keyword-Grounded Fact Substitution","display_name":"Mitigating Database Leakage in RAG Systems with Keyword-Grounded Fact Substitution","publication_year":2026,"publication_date":"2026-08-21","ids":{"openalex":"https://openalex.org/W7204154663","doi":"https://doi.org/10.48550/arxiv.2608.21656"},"language":null,"primary_location":{"id":"pmh:oai:arXiv.org:2608.21656","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.21656","pdf_url":"https://arxiv.org/pdf/2608.21656","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2608.21656","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5148062101","display_name":"Ziliang Zhang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Ziliang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5110958027","display_name":"Yubo Zhu","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhu, Yubo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5148132291","display_name":"Wei Tong","orcid":"https://orcid.org/0009-0004-5870-1499","observed_orcids":["https://orcid.org/0009-0004-5870-1499"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tong, Wei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101950654","display_name":"Jingyu Hua","orcid":"https://orcid.org/0000-0002-6709-4801","observed_orcids":["https://orcid.org/0000-0002-6709-4801"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hua, Jingyu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5140184588","display_name":"Zijian Wang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Zijian","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5148167828","display_name":"Yuan Zhang","orcid":"https://orcid.org/0000-0001-9682-5231","observed_orcids":["https://orcid.org/0000-0001-9682-5231"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Yuan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5127975537","display_name":"Sheng Zhong","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhong, Sheng","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":false,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.9955999851226807,"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":0.9955999851226807,"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/T10028","display_name":"Topic Modeling","score":0.003000000026077032,"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":0.0003000000142492354,"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/retrieval-augmented-generation","display_name":"retrieval-augmented generation","score":0.9860000014305115},{"id":"https://openalex.org/keywords/prompt-injection","display_name":"prompt injection","score":0.9739999771118164},{"id":"https://openalex.org/keywords/attention-rollout","display_name":"attention rollout","score":0.7599999904632568},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.5019999742507935}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7271999716758728},{"id":"https://openalex.org/C153701036","wikidata":"https://www.wikidata.org/wiki/Q659974","display_name":"Trustworthiness","level":2,"score":0.614300012588501},{"id":"https://openalex.org/C2779201187","wikidata":"https://www.wikidata.org/wiki/Q2775060","display_name":"Information leakage","level":2,"score":0.5383999943733215},{"id":"https://openalex.org/C2777042071","wikidata":"https://www.wikidata.org/wiki/Q6509304","display_name":"Leakage (economics)","level":2,"score":0.5218999981880188},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.47290000319480896},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.4505000114440918},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.445499986410141},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.4221999943256378},{"id":"https://openalex.org/C132964779","wikidata":"https://www.wikidata.org/wiki/Q2110223","display_name":"Raw data","level":2,"score":0.42160001397132874},{"id":"https://openalex.org/C77088390","wikidata":"https://www.wikidata.org/wiki/Q8513","display_name":"Database","level":1,"score":0.4011000096797943},{"id":"https://openalex.org/C2780992000","wikidata":"https://www.wikidata.org/wiki/Q17016113","display_name":"Generator (circuit theory)","level":3,"score":0.39320001006126404},{"id":"https://openalex.org/C2778744189","wikidata":"https://www.wikidata.org/wiki/Q4677551","display_name":"Active database","level":2,"score":0.29980000853538513},{"id":"https://openalex.org/C70061542","wikidata":"https://www.wikidata.org/wiki/Q989016","display_name":"Distributed database","level":2,"score":0.2858000099658966},{"id":"https://openalex.org/C137822555","wikidata":"https://www.wikidata.org/wiki/Q2587068","display_name":"Information sensitivity","level":2,"score":0.2806999981403351},{"id":"https://openalex.org/C112930515","wikidata":"https://www.wikidata.org/wiki/Q4389547","display_name":"Risk analysis (engineering)","level":1,"score":0.2766000032424927},{"id":"https://openalex.org/C2777168461","wikidata":"https://www.wikidata.org/wiki/Q42196253","display_name":"Set operations","level":3,"score":0.26190000772476196}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2608.21656","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.21656","pdf_url":"https://arxiv.org/pdf/2608.21656","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2608.21656","doi":"https://doi.org/10.48550/arxiv.2608.21656","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.21656","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2608.21656","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.21656","pdf_url":"https://arxiv.org/pdf/2608.21656","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"sustainable_development_goals_aurora":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Retrieval-Augmented":[0],"Generation":[1],"(RAG)":[2],"has":[3],"emerged":[4],"as":[5],"a":[6,47,63,78,94,152],"powerful":[7],"paradigm":[8],"for":[9],"combining":[10],"large":[11],"language":[12],"models":[13],"(LLMs)":[14],"with":[15,110],"external":[16],"knowledge":[17],"sources.":[18],"However,":[19],"RAG":[20,160],"systems":[21],"remain":[22],"vulnerable":[23],"to":[24,35,87,92],"prompt":[25],"injection":[26,141],"attacks,":[27],"which":[28],"may":[29],"mislead":[30],"the":[31,55,70,101,105,116,124,135],"retriever":[32],"or":[33],"generator":[34,117],"expose":[36],"sensitive":[37],"database":[38,138],"contents.":[39],"To":[40],"address":[41],"this":[42],"issue,":[43],"we":[44],"propose":[45],"KFS-RAG,":[46],"defense":[48],"that":[49,115,131],"mitigates":[50],"information":[51],"leakage":[52,139],"by":[53],"reformulating":[54],"retrieved":[56,71,102,126],"context.":[57],"Specifically,":[58],"our":[59],"method":[60],"first":[61],"identifies":[62],"small":[64],"set":[65,96],"of":[66,97,137],"influential":[67],"keywords":[68,83],"from":[69,100],"context":[72,107],"via":[73],"an":[74,89],"attention":[75],"rollout":[76],"plus":[77],"causal":[79],"perturbation":[80],"mechanism.":[81],"These":[82],"are":[84],"then":[85],"used":[86],"guide":[88],"auxiliary":[90],"LLM":[91],"generate":[93],"compact":[95],"keyword-grounded":[98],"facts":[99],"passages.":[103],"Finally,":[104],"original":[106],"is":[108],"substituted":[109],"these":[111],"curated":[112],"facts,":[113],"ensuring":[114],"operates":[118],"on":[119],"sanitized":[120],"evidence":[121],"rather":[122],"than":[123],"raw":[125],"text.":[127],"Experimental":[128],"evaluations":[129],"demonstrate":[130],"KFS-RAG":[132],"significantly":[133],"reduces":[134],"risk":[136],"under":[140],"attacks":[142],"while":[143],"maintaining":[144],"response":[145],"accuracy":[146],"and":[147,158],"relevance.":[148],"This":[149],"work":[150],"highlights":[151],"practical":[153],"pathway":[154],"toward":[155],"building":[156],"secure":[157],"trustworthy":[159],"systems.":[161]},"counts_by_year":[],"updated_date":"2026-10-01T08:54:57.152042","created_date":"2026-08-26T00:00:00"}