{"id":"https://openalex.org/W7161818990","doi":"https://doi.org/10.48550/arxiv.2605.18908","title":"Fast and Lightweight Backdoor Detection via Head Random Probing","display_name":"Fast and Lightweight Backdoor Detection via Head Random Probing","publication_year":2026,"publication_date":"2026-05-17","ids":{"openalex":"https://openalex.org/W7161818990","doi":"https://doi.org/10.48550/arxiv.2605.18908"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.18908","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.18908","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":"public-domain","license_id":"https://openalex.org/licenses/public-domain","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2605.18908","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5087622786","display_name":"Yinbo Yu","orcid":"https://orcid.org/0000-0002-0257-5081"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yu, Yinbo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136581499","display_name":"Xueyu Yin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yin, Xueyu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136533681","display_name":"Jing Fang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fang, Jing","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136551996","display_name":"Chunwei Tian","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tian, Chunwei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136555363","display_name":"Qi Zhu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhu, Qi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136544593","display_name":"Jiajia Liu","orcid":"https://orcid.org/0009-0003-8007-675X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Jiajia","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5136560699","display_name":"Daoqiang Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Daoqiang","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":0.9035000205039978,"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":0.9035000205039978,"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":0.021199999377131462,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.014700000174343586,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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":0.9825999736785889},{"id":"https://openalex.org/keywords/robustness","display_name":"Robustness (evolution)","score":0.7128000259399414},{"id":"https://openalex.org/keywords/detector","display_name":"Detector","score":0.4311000108718872},{"id":"https://openalex.org/keywords/benchmark","display_name":"Benchmark (surveying)","score":0.4115000069141388},{"id":"https://openalex.org/keywords/false-positive-rate","display_name":"False positive rate","score":0.3772999942302704},{"id":"https://openalex.org/keywords/artificial-neural-network","display_name":"Artificial neural network","score":0.3531000018119812}],"concepts":[{"id":"https://openalex.org/C2781045450","wikidata":"https://www.wikidata.org/wiki/Q254569","display_name":"Backdoor","level":2,"score":0.9825999736785889},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.7128000259399414},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.64410001039505},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4693000018596649},{"id":"https://openalex.org/C94915269","wikidata":"https://www.wikidata.org/wiki/Q1834857","display_name":"Detector","level":2,"score":0.4311000108718872},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4244000017642975},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.4115000069141388},{"id":"https://openalex.org/C95922358","wikidata":"https://www.wikidata.org/wiki/Q5432725","display_name":"False positive rate","level":2,"score":0.3772999942302704},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.3531000018119812},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.351500004529953},{"id":"https://openalex.org/C2984842247","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep neural networks","level":3,"score":0.3424000144004822},{"id":"https://openalex.org/C159694833","wikidata":"https://www.wikidata.org/wiki/Q2321565","display_name":"Iterative method","level":2,"score":0.26330000162124634},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.25429999828338623}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.18908","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.18908","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":"public-domain","license_id":"https://openalex.org/licenses/public-domain","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.18908","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.18908","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":"public-domain","license_id":"https://openalex.org/licenses/public-domain","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Deep":[0],"neural":[1],"networks":[2],"(DNNs)":[3],"remain":[4],"critically":[5],"vulnerable":[6],"to":[7,25,72],"backdoor":[8,47,144,192],"attacks.":[9,145],"Existing":[10],"post-training":[11],"detectors":[12],"often":[13],"require":[14],"clean":[15,133],"or":[16,20,109,114],"surrogate":[17,110],"data,":[18,111],"gradients,":[19,113],"iterative":[21],"trigger":[22,58],"reconstruction,":[23],"leading":[24],"high":[26],"computational":[27],"costs":[28],"and":[29,44,98,130,140,152,186],"limited":[30],"robustness":[31],"under":[32,81],"practical":[33],"model-auditing":[34],"scenarios.":[35],"In":[36],"this":[37],"paper,":[38],"we":[39],"propose":[40],"HTell,":[41],"a":[42,121],"fast":[43],"lightweight":[45],"data-free":[46,191],"detector":[48],"based":[49],"on":[50,77,120],"head":[51,179],"random":[52,82,88,180],"probing.":[53],"Instead":[54],"of":[55,143],"reconstructing":[56],"diverse":[57],"patterns,":[59],"HTell":[60,85,119,146],"inspects":[61],"their":[62],"unified":[63],"manifestation":[64],"in":[65],"the":[66,78,95,164],"prediction":[67],"head:":[68],"backdoored":[69,128],"models":[70,129],"tend":[71],"exhibit":[73],"abnormal":[74],"response":[75,104],"concentration":[76],"target":[79],"class":[80],"latent":[83,89],"probes.":[84],"generates":[86],"architecture-aware":[87],"probes,":[90],"feeds":[91],"them":[92],"directly":[93],"into":[94],"model":[96,112,193],"head,":[97],"detects":[99],"backdoors":[100],"by":[101,167],"analyzing":[102],"class-wise":[103],"statistics,":[105],"without":[106],"accessing":[107],"real":[108],"parameter":[115],"optimization.":[116],"We":[117],"evaluate":[118],"large-scale":[122,190],"benchmark":[123],"containing":[124],"more":[125],"than":[126],"6,000":[127],"over":[131,168],"700":[132],"models,":[134],"covering":[135],"4":[136],"datasets,":[137],"14":[138],"architectures,":[139],"21":[141],"types":[142],"achieves":[147],"99.03%":[148],"true":[149],"positive":[150,155],"rate":[151,156],"2.11%":[153],"false":[154],"with":[157,171],"only":[158],"12.69":[159],"ms/model":[160],"detection":[161],"latency,":[162],"reducing":[163],"time":[165],"cost":[166],"30,000$\\times$":[169],"compared":[170],"representative":[172],"gradient-based":[173],"detectors.":[174],"These":[175],"results":[176],"demonstrate":[177],"that":[178],"probing":[181],"provides":[182],"an":[183],"accurate,":[184],"robust,":[185],"efficient":[187],"solution":[188],"for":[189],"auditing.":[194]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-21T00:00:00"}