{"id":"https://openalex.org/W4416251398","doi":"https://doi.org/10.1109/ijcnn64981.2025.11228642","title":"Less is Better: Exploiting Label Bias to Enhance Instruction Tuning-based Backdoor Attack on Large Language Models","display_name":"Less is Better: Exploiting Label Bias to Enhance Instruction Tuning-based Backdoor Attack on Large Language Models","publication_year":2025,"publication_date":"2025-06-30","ids":{"openalex":"https://openalex.org/W4416251398","doi":"https://doi.org/10.1109/ijcnn64981.2025.11228642"},"language":null,"primary_location":{"id":"doi:10.1109/ijcnn64981.2025.11228642","is_oa":false,"landing_page_url":"https://doi.org/10.1109/ijcnn64981.2025.11228642","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 International Joint Conference on Neural Networks (IJCNN)","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5014537056","display_name":"Hengyang Lu","orcid":"https://orcid.org/0000-0001-5321-705X"},"institutions":[{"id":"https://openalex.org/I111599522","display_name":"Jiangnan University","ror":"https://ror.org/04mkzax54","country_code":"CN","type":"education","lineage":["https://openalex.org/I111599522"]}],"countries":["CN"],"is_corresponding":true,"raw_author_name":"Heng-Yang Lu","raw_affiliation_strings":["Jiangnan University,School of Artificial Intelligence and Computer Science,Wuxi,China"],"affiliations":[{"raw_affiliation_string":"Jiangnan University,School of Artificial Intelligence and Computer Science,Wuxi,China","institution_ids":["https://openalex.org/I111599522"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5120499605","display_name":"Yi Zhong","orcid":null},"institutions":[{"id":"https://openalex.org/I111599522","display_name":"Jiangnan University","ror":"https://ror.org/04mkzax54","country_code":"CN","type":"education","lineage":["https://openalex.org/I111599522"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yi Zhong","raw_affiliation_strings":["Jiangnan University,School of Artificial Intelligence and Computer Science,Wuxi,China"],"affiliations":[{"raw_affiliation_string":"Jiangnan University,School of Artificial Intelligence and Computer Science,Wuxi,China","institution_ids":["https://openalex.org/I111599522"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5073738793","display_name":"Changrong Xia","orcid":"https://orcid.org/0000-0002-4254-1425"},"institutions":[{"id":"https://openalex.org/I143868143","display_name":"Anhui University","ror":"https://ror.org/05th6yx34","country_code":"CN","type":"education","lineage":["https://openalex.org/I143868143"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Chang Xia","raw_affiliation_strings":["Anhui University,School of Computer Science and Technology,Anhui,China"],"affiliations":[{"raw_affiliation_string":"Anhui University,School of Computer Science and Technology,Anhui,China","institution_ids":["https://openalex.org/I143868143"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":["https://openalex.org/A5014537056"],"corresponding_institution_ids":["https://openalex.org/I111599522"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.18277086,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"8"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.9503999948501587,"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":0.9503999948501587,"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/T12262","display_name":"Hate Speech and Cyberbullying Detection","score":0.007199999876320362,"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":0.006800000090152025,"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/backdoor","display_name":"Backdoor","score":0.9926999807357788},{"id":"https://openalex.org/keywords/vulnerability","display_name":"Vulnerability (computing)","score":0.5782999992370605},{"id":"https://openalex.org/keywords/set","display_name":"Set (abstract data type)","score":0.5246999859809875},{"id":"https://openalex.org/keywords/selection","display_name":"Selection (genetic algorithm)","score":0.4302999973297119},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.4075999855995178}],"concepts":[{"id":"https://openalex.org/C2781045450","wikidata":"https://www.wikidata.org/wiki/Q254569","display_name":"Backdoor","level":2,"score":0.9926999807357788},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7566999793052673},{"id":"https://openalex.org/C95713431","wikidata":"https://www.wikidata.org/wiki/Q631425","display_name":"Vulnerability (computing)","level":2,"score":0.5782999992370605},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.5246999859809875},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.4302999973297119},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.4075999855995178},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3675000071525574},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.35370001196861267},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.33550000190734863},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.33180001378059387},{"id":"https://openalex.org/C40423286","wikidata":"https://www.wikidata.org/wiki/Q284172","display_name":"Selection bias","level":2,"score":0.31369999051094055},{"id":"https://openalex.org/C58489278","wikidata":"https://www.wikidata.org/wiki/Q1172284","display_name":"Data set","level":2,"score":0.2502000033855438}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/ijcnn64981.2025.11228642","is_oa":false,"landing_page_url":"https://doi.org/10.1109/ijcnn64981.2025.11228642","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 International Joint Conference on Neural Networks (IJCNN)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320321543","display_name":"China Postdoctoral Science Foundation","ror":"https://ror.org/0426zh255"}],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":8,"referenced_works":["https://openalex.org/W2973217491","https://openalex.org/W2996851481","https://openalex.org/W3109409894","https://openalex.org/W3176270593","https://openalex.org/W3196832521","https://openalex.org/W4391696889","https://openalex.org/W4401042536","https://openalex.org/W4401042878"],"related_works":[],"abstract_inverted_index":{"Recent":[0],"research":[1],"has":[2],"uncovered":[3],"the":[4,37,42,70,75,79,89,102,136,157,177,188],"vulnerability":[5],"of":[6,77,138],"Large":[7],"Language":[8],"Model":[9,146],"(LLM)":[10],"instruction":[11,22,27,34,133],"tuning":[12],"via":[13],"BackDoor":[14],"Attacks":[15],"(BDA).":[16],"Given":[17],"a":[18,52,59,116],"poisoned":[19,33,67,108,121],"LLM":[20],"after":[21],"tuning,":[23,134],"input":[24,31],"with":[25,32,119,183,190],"clean":[26],"performs":[28],"normally":[29],"while":[30],"would":[35],"give":[36],"target":[38,94,103],"response":[39],"set":[40],"by":[41],"attacker.":[43],"We":[44,81,123,149],"have":[45],"witnessed":[46],"that":[47,83,171],"existing":[48],"works":[49],"usually":[50],"require":[51],"5%":[53],"Poisoning":[54],"Ratio":[55],"(PR)":[56],"to":[57,92,106,114,181,197],"achieve":[58,115,176],"high":[60,117],"Attack":[61],"Success":[62],"Rate":[63],"(ASR).":[64],"However,":[65],"more":[66],"samples":[68],"in":[69,97],"training":[71],"data":[72],"may":[73],"increase":[74],"risk":[76],"exposing":[78],"BDA.":[80],"observe":[82],"using":[84],"certain":[85],"specific":[86],"instructions":[87,109],"makes":[88],"model":[90],"tend":[91],"predict":[93],"labels.":[95],"Thus,":[96],"this":[98],"work,":[99],"we":[100],"utilize":[101],"label":[104],"bias":[105],"generate":[107],"as":[110,113],"triggers,":[111],"so":[112],"ASR":[118,179,192],"less":[120],"data.":[122],"propose":[124],"Label":[125],"Bias":[126,139,142],"Enhanced":[127],"(LBE)":[128],"backdoor":[129],"attacks":[130],"based":[131],"on":[132,152,160],"including":[135],"module":[137],"Instruction":[140,143],"Generation,":[141],"Selection":[144],"and":[145,155,167],"Contamination":[147],"Module.":[148],"perform":[150],"experiments":[151],"three":[153,161],"datasets":[154],"evaluate":[156],"BDA":[158],"performance":[159],"popular":[162],"open-sourced":[163],"LLMs":[164],"(Llama2,":[165],"ChatGLM3":[166],"Vicuna).":[168],"Results":[169],"show":[170],"our":[172],"proposed":[173],"LBE":[174],"can":[175],"highest":[178],"(close":[180],"100%)":[182],"1%":[184],"poisoning":[185],"ratio,":[186],"exceeding":[187],"baselines":[189],"an":[191],"improvement":[193],"ranging":[194],"from":[195],"2%":[196],"20%.":[198]},"counts_by_year":[],"updated_date":"2026-04-09T08:11:56.329763","created_date":"2025-11-14T00:00:00"}