{"id":"https://openalex.org/W7147709814","doi":"https://doi.org/10.48550/arxiv.2603.28673","title":"FL-PBM: Pre-Training Backdoor Mitigation for Federated Learning","display_name":"FL-PBM: Pre-Training Backdoor Mitigation for Federated Learning","publication_year":2026,"publication_date":"2026-03-30","ids":{"openalex":"https://openalex.org/W7147709814","doi":"https://doi.org/10.48550/arxiv.2603.28673"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2603.28673","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.28673","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":null,"license_id":null,"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.2603.28673","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5001753157","display_name":"Osama Wehbi","orcid":"https://orcid.org/0000-0002-2121-218X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wehbi, Osama","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5029454859","display_name":"Sarhad Arisdakessian","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Arisdakessian, Sarhad","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5069036268","display_name":"Omar Abdel Wahab","orcid":"https://orcid.org/0000-0002-3991-4673"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wahab, Omar Abdel","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5005269467","display_name":"Azzam Mourad","orcid":"https://orcid.org/0000-0001-9434-5322"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mourad, Azzam","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5049846737","display_name":"Hadi Otrok","orcid":"https://orcid.org/0000-0002-9574-5384"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Otrok, Hadi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5120968890","display_name":"Jamal Bentahar","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bentahar, Jamal","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":0.9717000126838684,"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":0.9717000126838684,"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/T10764","display_name":"Privacy-Preserving Technologies in Data","score":0.007300000172108412,"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":0.004399999976158142,"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":0.996999979019165},{"id":"https://openalex.org/keywords/discriminative-model","display_name":"Discriminative model","score":0.5609999895095825},{"id":"https://openalex.org/keywords/cluster-analysis","display_name":"Cluster analysis","score":0.5562999844551086},{"id":"https://openalex.org/keywords/federated-learning","display_name":"Federated learning","score":0.45239999890327454},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.4489000141620636},{"id":"https://openalex.org/keywords/reliability","display_name":"Reliability (semiconductor)","score":0.426800012588501}],"concepts":[{"id":"https://openalex.org/C2781045450","wikidata":"https://www.wikidata.org/wiki/Q254569","display_name":"Backdoor","level":2,"score":0.996999979019165},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.722599983215332},{"id":"https://openalex.org/C97931131","wikidata":"https://www.wikidata.org/wiki/Q5282087","display_name":"Discriminative model","level":2,"score":0.5609999895095825},{"id":"https://openalex.org/C73555534","wikidata":"https://www.wikidata.org/wiki/Q622825","display_name":"Cluster analysis","level":2,"score":0.5562999844551086},{"id":"https://openalex.org/C2992525071","wikidata":"https://www.wikidata.org/wiki/Q50818671","display_name":"Federated learning","level":2,"score":0.45239999890327454},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.4489000141620636},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.43709999322891235},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.4345000088214874},{"id":"https://openalex.org/C43214815","wikidata":"https://www.wikidata.org/wiki/Q7310987","display_name":"Reliability (semiconductor)","level":3,"score":0.426800012588501},{"id":"https://openalex.org/C168167062","wikidata":"https://www.wikidata.org/wiki/Q1117970","display_name":"Component (thermodynamics)","level":2,"score":0.4133000075817108},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3612000048160553},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.36000001430511475},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.3472999930381775},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.32260000705718994},{"id":"https://openalex.org/C12725497","wikidata":"https://www.wikidata.org/wiki/Q810247","display_name":"Baseline (sea)","level":2,"score":0.31470000743865967},{"id":"https://openalex.org/C61224824","wikidata":"https://www.wikidata.org/wiki/Q2260434","display_name":"Mixture model","level":2,"score":0.31189998984336853}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2603.28673","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.28673","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2603.28673","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.28673","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/10","score":0.4479488730430603,"display_name":"Reduced inequalities"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Backdoor":[0,71],"attacks":[1,30,51],"pose":[2],"a":[3,77,94,107,115,159],"significant":[4],"threat":[5],"to":[6,18,33,113,124,142,163,206,209,217,220],"the":[7,55,87,111,130,133,153,183,189,227],"integrity":[8],"and":[9,45,48,63,128,156,178,214,224],"reliability":[10],"of":[11,57,102,132,185],"Artificial":[12],"Intelligence":[13],"(AI)":[14],"models,":[15],"enabling":[16],"adversaries":[17],"manipulate":[19],"model":[20,91,235,245],"behavior":[21],"by":[22,204,215],"injecting":[23],"poisoned":[24,84],"data":[25,85,112,146,174],"with":[26],"hidden":[27],"triggers.":[28,167],"These":[29],"can":[31],"lead":[32],"severe":[34],"consequences,":[35],"especially":[36],"in":[37,93,152,237],"critical":[38],"applications":[39],"such":[40],"as":[41],"autonomous":[42],"driving,":[43],"healthcare,":[44],"finance.":[46],"Detecting":[47],"mitigating":[49],"backdoor":[50,166,186],"is":[52,175],"crucial":[53],"across":[54],"lifespan":[56],"model's":[58],"phases,":[59],"including":[60],"pre-training,":[61],"in-training,":[62],"post-training.":[64],"In":[65],"this":[66],"paper,":[67],"we":[68],"propose":[69],"Pre-Training":[70],"Mitigation":[72],"for":[73],"Federated":[74],"Learning":[75],"(FL-PBM),":[76],"novel":[78],"defense":[79],"mechanism":[80],"that":[81,172,198],"proactively":[82],"filters":[83],"on":[86,149,188,194],"client":[88],"side":[89],"before":[90],"training":[92],"federated":[95,211],"learning":[96,212],"(FL)":[97],"environment.":[98],"The":[99],"approach":[100],"consists":[101],"three":[103],"stages:":[104],"(1)":[105],"inserting":[106],"benign":[108],"trigger":[109],"into":[110],"establish":[114],"controlled":[116],"baseline,":[117],"(2)":[118],"applying":[119,158],"Principal":[120],"Component":[121],"Analysis":[122],"(PCA)":[123],"extract":[125],"discriminative":[126],"features":[127],"assess":[129],"separability":[131],"data,":[134],"(3)":[135],"performing":[136],"Gaussian":[137],"Mixture":[138],"Model":[139],"(GMM)":[140],"clustering":[141],"identify":[143],"potentially":[144],"malicious":[145],"samples":[147],"based":[148],"their":[150],"distribution":[151],"PCA-transformed":[154],"space,":[155],"(4)":[157],"targeted":[160],"blurring":[161],"technique":[162],"disrupt":[164],"potential":[165],"Together,":[168],"these":[169],"steps":[170],"ensure":[171],"suspicious":[173],"detected":[176],"early":[177],"sanitized":[179],"effectively,":[180],"thereby":[181],"minimizing":[182],"influence":[184],"triggers":[187],"global":[190],"model.":[191],"Experimental":[192],"evaluations":[193],"image-based":[195],"datasets":[196],"demonstrate":[197],"FL-PBM":[199],"reduces":[200],"attack":[201],"success":[202],"rates":[203],"up":[205],"95%":[207],"compared":[208],"baseline":[210],"(FedAvg)":[213],"30":[216],"80%":[218],"relative":[219],"state-of-the-art":[221],"defenses":[222],"(RDFL":[223],"LPSF).":[225],"At":[226],"same":[228],"time,":[229],"it":[230],"maintains":[231],"over":[232],"90%":[233],"clean":[234],"accuracy":[236],"most":[238],"experiments,":[239],"achieving":[240],"better":[241],"mitigation":[242],"without":[243],"degrading":[244],"performance.":[246]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-04-02T00:00:00"}