{"id":"https://openalex.org/W4411120592","doi":"https://doi.org/10.18653/v1/2025.findings-naacl.316","title":"Data-centric NLP Backdoor Defense from the Lens of Memorization","display_name":"Data-centric NLP Backdoor Defense from the Lens of Memorization","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4411120592","doi":"https://doi.org/10.18653/v1/2025.findings-naacl.316"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.findings-naacl.316","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-naacl.316","pdf_url":"https://aclanthology.org/2025.findings-naacl.316.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: NAACL 2025","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.findings-naacl.316.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5044589781","display_name":"Zhenting Wang","orcid":"https://orcid.org/0000-0001-6527-5089"},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Zhenting Wang","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101850444","display_name":"Zhizhi Wang","orcid":"https://orcid.org/0000-0003-2223-9621"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhizhi Wang","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102521857","display_name":"Mingyu Jin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mingyu Jin","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5072191151","display_name":"Mengnan Du","orcid":"https://orcid.org/0000-0002-1614-6069"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mengnan Du","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5071575216","display_name":"Juan Zhai","orcid":"https://orcid.org/0000-0001-5017-8016"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Juan Zhai","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5028714018","display_name":"Shiqing Ma","orcid":"https://orcid.org/0000-0003-4741-7674"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shiqing Ma","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":6,"corresponding_author_ids":["https://openalex.org/A5044589781"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.05984188,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"5713","last_page":"5731"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.991599977016449,"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":0.991599977016449,"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":0.9898999929428101,"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/T10400","display_name":"Network Security and Intrusion Detection","score":0.9617000222206116,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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":0.9643046259880066},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.6674354076385498},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.6487532258033752},{"id":"https://openalex.org/keywords/memorization","display_name":"Memorization","score":0.6424846053123474},{"id":"https://openalex.org/keywords/natural-language-processing","display_name":"Natural language processing","score":0.6204500198364258},{"id":"https://openalex.org/keywords/speech-recognition","display_name":"Speech recognition","score":0.3201230764389038},{"id":"https://openalex.org/keywords/linguistics","display_name":"Linguistics","score":0.17131027579307556},{"id":"https://openalex.org/keywords/computer-security","display_name":"Computer security","score":0.14020997285842896},{"id":"https://openalex.org/keywords/philosophy","display_name":"Philosophy","score":0.10833403468132019}],"concepts":[{"id":"https://openalex.org/C2781045450","wikidata":"https://www.wikidata.org/wiki/Q254569","display_name":"Backdoor","level":2,"score":0.9643046259880066},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6674354076385498},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6487532258033752},{"id":"https://openalex.org/C30038468","wikidata":"https://www.wikidata.org/wiki/Q4354775","display_name":"Memorization","level":2,"score":0.6424846053123474},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.6204500198364258},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.3201230764389038},{"id":"https://openalex.org/C41895202","wikidata":"https://www.wikidata.org/wiki/Q8162","display_name":"Linguistics","level":1,"score":0.17131027579307556},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.14020997285842896},{"id":"https://openalex.org/C138885662","wikidata":"https://www.wikidata.org/wiki/Q5891","display_name":"Philosophy","level":0,"score":0.10833403468132019}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.findings-naacl.316","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-naacl.316","pdf_url":"https://aclanthology.org/2025.findings-naacl.316.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: NAACL 2025","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.findings-naacl.316","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-naacl.316","pdf_url":"https://aclanthology.org/2025.findings-naacl.316.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: NAACL 2025","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G175455262","display_name":null,"funder_award_id":"W911NF-19-S-0012","funder_id":"https://openalex.org/F4320333051","funder_display_name":"Intelligence Advanced Research Projects Activity"},{"id":"https://openalex.org/G1865040260","display_name":null,"funder_award_id":"TrojAI W911NF-19-S-0012","funder_id":"https://openalex.org/F4320333051","funder_display_name":"Intelligence Advanced Research Projects Activity"},{"id":"https://openalex.org/G3763956660","display_name":null,"funder_award_id":"2319944","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G608057831","display_name":null,"funder_award_id":"W911NF-19","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G6897929834","display_name":null,"funder_award_id":"2342250","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"},{"id":"https://openalex.org/F4320333051","display_name":"Intelligence Advanced Research Projects Activity","ror":"https://ror.org/01v3fsc55"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4411120592.pdf","grobid_xml":"https://content.openalex.org/works/W4411120592.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4320031223","https://openalex.org/W4200629851","https://openalex.org/W4281902577","https://openalex.org/W4309417370","https://openalex.org/W4292107232","https://openalex.org/W3009072493","https://openalex.org/W4386080799","https://openalex.org/W3140988292","https://openalex.org/W4317672133","https://openalex.org/W3204019825"],"abstract_inverted_index":{"Backdoor":[0],"attack":[1],"is":[2,62],"a":[3,47,89],"severe":[4],"threat":[5],"to":[6,27,65],"the":[7,18,57,66,72,114],"trustworthiness":[8],"of":[9,20,22,49,59,68,135],"DNN-based":[10],"language":[11,23,43],"models.In":[12],"this":[13],"paper,":[14],"we":[15,54,87],"first":[16,92],"extend":[17],"definition":[19],"memorization":[21,61],"models":[24],"from":[25],"samplewise":[26],"more":[28],"fine-grained":[29],"sentence":[30,77],"elementwise":[31],"(e.g.,":[32],"word,":[33],"phrase,":[34],"structure,":[35],"and":[36,38,106],"style),":[37],"then":[39,107],"point":[40],"out":[41],"that":[42,56,124],"model":[44],"backdoors":[45],"are":[46,79],"type":[48],"element-wise":[50],"memorization.Through":[51],"further":[52],"analysis,":[53],"find":[55],"strength":[58],"such":[60],"positively":[63],"correlated":[64],"frequency":[67],"duplicated":[69,76,104],"elements":[70,78],"in":[71,96,130],"training":[73,97],"dataset.In":[74],"conclusion,":[75],"necessary":[80],"for":[81],"successful":[82],"backdoor":[83,118],"attacks.Based":[84],"on":[85],"this,":[86],"propose":[88],"data-centric":[90],"defense.We":[91],"detect":[93],"trigger":[94],"candidates":[95,115],"data":[98],"by":[99,111],"finding":[100],"memorizable":[101],"elements,":[102,105],"i.e.,":[103],"confirm":[108],"real":[109],"triggers":[110],"testing":[112],"if":[113],"can":[116],"activate":[117],"behaviors":[119],"(i.e.,":[120],"malicious":[121],"elements).Results":[122],"show":[123],"our":[125],"method":[126],"outperforms":[127],"state-of-the-art":[128],"defenses":[129],"defending":[131],"against":[132],"different":[133],"types":[134],"NLP":[136],"backdoors.":[137]},"counts_by_year":[],"updated_date":"2026-04-21T08:09:41.155169","created_date":"2025-10-10T00:00:00"}