{"id":"https://openalex.org/W7161181276","doi":"https://doi.org/10.48550/arxiv.2605.13418","title":"DP-KFC: Data-Free Preconditioning for Privacy-Preserving Deep Learning","display_name":"DP-KFC: Data-Free Preconditioning for Privacy-Preserving Deep Learning","publication_year":2026,"publication_date":"2026-05-13","ids":{"openalex":"https://openalex.org/W7161181276","doi":"https://doi.org/10.48550/arxiv.2605.13418"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.13418","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.13418","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","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.2605.13418","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5136102326","display_name":"Marc Molina Van den Bosch","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bosch, Marc Molina Van den","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5050084525","display_name":"Riccardo Taiello","orcid":"https://orcid.org/0000-0002-9890-9639"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Taiello, Riccardo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5095787326","display_name":"Albert Sund Aillet","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Aillet, Albert Sund","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5111396570","display_name":"Andrea Protani","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Protani, Andrea","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136185588","display_name":"Miguel Angel Gonzalez Ballester","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ballester, Miguel Angel Gonzalez","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5069214353","display_name":"Luigi Serio","orcid":"https://orcid.org/0000-0002-9346-2663"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Serio, Luigi","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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.3797000050544739,"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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.3797000050544739,"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":0.3303000032901764,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.12720000743865967,"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/differential-privacy","display_name":"Differential privacy","score":0.5397999882698059},{"id":"https://openalex.org/keywords/deep-learning","display_name":"Deep learning","score":0.5335999727249146},{"id":"https://openalex.org/keywords/synthetic-data","display_name":"Synthetic data","score":0.5116999745368958},{"id":"https://openalex.org/keywords/modalities","display_name":"Modalities","score":0.5102999806404114},{"id":"https://openalex.org/keywords/fisher-information","display_name":"Fisher information","score":0.5026999711990356},{"id":"https://openalex.org/keywords/information-privacy","display_name":"Information privacy","score":0.489300012588501},{"id":"https://openalex.org/keywords/curvature","display_name":"Curvature","score":0.4239000082015991},{"id":"https://openalex.org/keywords/distribution","display_name":"Distribution (mathematics)","score":0.3619000017642975}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6851000189781189},{"id":"https://openalex.org/C23130292","wikidata":"https://www.wikidata.org/wiki/Q5275358","display_name":"Differential privacy","level":2,"score":0.5397999882698059},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.5335999727249146},{"id":"https://openalex.org/C160920958","wikidata":"https://www.wikidata.org/wiki/Q7662746","display_name":"Synthetic data","level":2,"score":0.5116999745368958},{"id":"https://openalex.org/C2779903281","wikidata":"https://www.wikidata.org/wiki/Q6888026","display_name":"Modalities","level":2,"score":0.5102999806404114},{"id":"https://openalex.org/C29406490","wikidata":"https://www.wikidata.org/wiki/Q1420659","display_name":"Fisher information","level":2,"score":0.5026999711990356},{"id":"https://openalex.org/C123201435","wikidata":"https://www.wikidata.org/wiki/Q456632","display_name":"Information privacy","level":2,"score":0.489300012588501},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.45899999141693115},{"id":"https://openalex.org/C195065555","wikidata":"https://www.wikidata.org/wiki/Q214881","display_name":"Curvature","level":2,"score":0.4239000082015991},{"id":"https://openalex.org/C110121322","wikidata":"https://www.wikidata.org/wiki/Q865811","display_name":"Distribution (mathematics)","level":2,"score":0.3619000017642975},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.36010000109672546},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.3456999957561493},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.3434999883174896},{"id":"https://openalex.org/C185429906","wikidata":"https://www.wikidata.org/wiki/Q1130160","display_name":"Estimator","level":2,"score":0.3093999922275543},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.30480000376701355},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.30410000681877136},{"id":"https://openalex.org/C184050105","wikidata":"https://www.wikidata.org/wiki/Q273163","display_name":"Isotropy","level":2,"score":0.2971000075340271},{"id":"https://openalex.org/C106487976","wikidata":"https://www.wikidata.org/wiki/Q685816","display_name":"Matrix (chemical analysis)","level":2,"score":0.29249998927116394},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.27480000257492065},{"id":"https://openalex.org/C137822555","wikidata":"https://www.wikidata.org/wiki/Q2587068","display_name":"Information sensitivity","level":2,"score":0.25679999589920044},{"id":"https://openalex.org/C21080849","wikidata":"https://www.wikidata.org/wiki/Q13611879","display_name":"Data point","level":2,"score":0.25279998779296875},{"id":"https://openalex.org/C8505890","wikidata":"https://www.wikidata.org/wiki/Q605095","display_name":"Budget constraint","level":2,"score":0.25099998712539673},{"id":"https://openalex.org/C99221444","wikidata":"https://www.wikidata.org/wiki/Q1532069","display_name":"Private information retrieval","level":2,"score":0.25029999017715454}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.13418","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.13418","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.13418","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.13418","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Differentially":[0],"private":[1,31,81],"optimization":[2],"suffers":[3],"from":[4,61],"a":[5],"fundamental":[6],"geometric":[7],"mismatch:":[8],"deep":[9],"networks":[10,74],"have":[11],"highly":[12],"anisotropic":[13],"loss":[14],"landscapes,":[15],"yet":[16],"DP-SGD":[17,89],"injects":[18],"isotropic":[19],"noise.":[20],"Second-order":[21],"preconditioning":[22],"can":[23,118],"resolve":[24],"this,":[25],"but":[26],"estimating":[27],"curvature":[28,117],"typically":[29],"requires":[30],"data":[32,38,143],"(consuming":[33],"privacy":[34,98,123],"budget)":[35],"or":[36,125],"public":[37,83],"(introducing":[39],"distribution":[40,127],"shift).":[41],"We":[42,65],"show":[43],"that":[44,116],"the":[45],"Fisher":[46],"Information":[47],"Matrix":[48],"decouples":[49],"into":[50],"architectural":[51],"sensitivity,":[52],"recoverable":[53],"via":[54],"synthetic":[55,77],"noise,":[56,78],"and":[57,90],"input":[58],"correlations,":[59],"approximable":[60],"modality-specific":[62],"frequency":[63],"statistics.":[64],"propose":[66],"DP-KFC,":[67],"which":[68],"constructs":[69],"KFAC":[70],"preconditioners":[71,106],"by":[72,111],"probing":[73],"with":[75],"structured":[76],"requiring":[79],"neither":[80],"nor":[82],"data.":[84],"Empirically,":[85],"DP-KFC":[86,103],"consistently":[87],"outperforms":[88],"adaptive":[91],"baselines":[92],"across":[93],"diverse":[94],"modalities":[95],"in":[96,133],"strong":[97],"regimes":[99],"($\\varepsilon":[100],"\\leq":[101],"3$).":[102],"matches":[104],"private-data":[105],"while":[107],"public-data":[108],"variants":[109],"degrade":[110],"up":[112],"to":[113],"$4.8\\%$,":[114],"showing":[115],"be":[119],"estimated":[120],"without":[121],"consuming":[122],"budget":[124],"introducing":[126],"shift.":[128],"This":[129],"enables":[130],"privacy-preserving":[131],"learning":[132],"specialized":[134],"domains":[135],"(e.g.,":[136],"medical":[137],"applications)":[138],"where":[139],"regulatory":[140],"constraints":[141],"make":[142],"scarce.":[144]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-15T00:00:00"}