{"id":"https://openalex.org/W7131386371","doi":"https://doi.org/10.48550/arxiv.2602.18758","title":"UFO: Unlocking Ultra-Efficient Quantized Private Inference with Protocol and Algorithm Co-Optimization","display_name":"UFO: Unlocking Ultra-Efficient Quantized Private Inference with Protocol and Algorithm Co-Optimization","publication_year":2026,"publication_date":"2026-02-21","ids":{"openalex":"https://openalex.org/W7131386371","doi":"https://doi.org/10.48550/arxiv.2602.18758"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2602.18758","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2602.18758","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.2602.18758","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5084065765","display_name":"Wenxuan Zeng","orcid":"https://orcid.org/0000-0003-3743-0947"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zeng, Wenxuan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126803206","display_name":"Chao Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Chao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126810737","display_name":"Tianshi Xu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Tianshi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126801468","display_name":"Bo Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Bo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126706096","display_name":"Changrui Ren","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ren, Changrui","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126827355","display_name":"Jin Song Dong","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dong, Jin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5126792729","display_name":"Meng Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Meng","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/T10764","display_name":"Privacy-Preserving Technologies in Data","score":0.43230000138282776,"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/T10764","display_name":"Privacy-Preserving Technologies in Data","score":0.43230000138282776,"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":0.19300000369548798,"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/T10237","display_name":"Cryptography and Data Security","score":0.08070000261068344,"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/inference","display_name":"Inference","score":0.7117000222206116},{"id":"https://openalex.org/keywords/quantization","display_name":"Quantization (signal processing)","score":0.6764000058174133},{"id":"https://openalex.org/keywords/convolutional-neural-network","display_name":"Convolutional neural network","score":0.47850000858306885},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.4578999876976013},{"id":"https://openalex.org/keywords/convolution","display_name":"Convolution (computer science)","score":0.44749999046325684},{"id":"https://openalex.org/keywords/protocol","display_name":"Protocol (science)","score":0.3882000148296356},{"id":"https://openalex.org/keywords/communications-protocol","display_name":"Communications protocol","score":0.38109999895095825}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7289999723434448},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.7117000222206116},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.6764000058174133},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.5577999949455261},{"id":"https://openalex.org/C81363708","wikidata":"https://www.wikidata.org/wiki/Q17084460","display_name":"Convolutional neural network","level":2,"score":0.47850000858306885},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.4578999876976013},{"id":"https://openalex.org/C45347329","wikidata":"https://www.wikidata.org/wiki/Q5166604","display_name":"Convolution (computer science)","level":3,"score":0.44749999046325684},{"id":"https://openalex.org/C2780385302","wikidata":"https://www.wikidata.org/wiki/Q367158","display_name":"Protocol (science)","level":3,"score":0.3882000148296356},{"id":"https://openalex.org/C12269588","wikidata":"https://www.wikidata.org/wiki/Q132364","display_name":"Communications protocol","level":2,"score":0.38109999895095825},{"id":"https://openalex.org/C2777472644","wikidata":"https://www.wikidata.org/wiki/Q16968992","display_name":"Approximate inference","level":3,"score":0.3727000057697296},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.3628999888896942},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.3515999913215637},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.34470000863075256},{"id":"https://openalex.org/C157899210","wikidata":"https://www.wikidata.org/wiki/Q1395022","display_name":"Convolutional code","level":3,"score":0.3357999920845032},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.29789999127388},{"id":"https://openalex.org/C74193536","wikidata":"https://www.wikidata.org/wiki/Q574844","display_name":"Kernel (algebra)","level":2,"score":0.2809000015258789},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.2759000062942505}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2602.18758","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2602.18758","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.2602.18758","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2602.18758","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":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Private":[0],"convolutional":[1],"neural":[2],"network":[3],"(CNN)":[4],"inference":[5,32,61,80,95,147],"based":[6,163],"on":[7,164],"secure":[8],"two-party":[9],"computation":[10],"(2PC)":[11],"suffers":[12],"from":[13,20],"high":[14],"communication":[15,100,172,205],"and":[16,40,70,87,131,203,210,220],"latency":[17],"overhead,":[18],"especially":[19],"convolution":[21,55,72],"layers.":[22],"In":[23],"this":[24],"paper,":[25],"we":[26,64,127,138,157,178],"propose":[27,139],"UFO,":[28],"a":[29,45,140,159,181],"quantized":[30],"2PC":[31,38,47,146],"framework":[33],"that":[34,49,66,112],"jointly":[35],"optimizes":[36],"the":[37,52,74,79,104,135,150,154,176,188],"protocols":[39],"quantization":[41,58,69],"algorithm.":[42,132],"UFO":[43,199],"features":[44],"novel":[46],"protocol":[48,130,136],"systematically":[50],"combines":[51],"efficient":[53],"Winograd":[54,71,82,107],"algorithm":[56,155,162,185],"with":[57,207],"to":[59,93,98,148,167,186,215],"improve":[60],"efficiency.":[62],"However,":[63],"observe":[65],"naively":[67],"combining":[68],"faces":[73],"following":[75],"challenges:":[76],"1)":[77,133],"From":[78,103],"perspective,":[81,106],"transformations":[83,108],"introduce":[84,109,180],"extensive":[85,197],"additions":[86],"require":[88],"frequent":[89],"bit":[90,183,194],"width":[91],"conversions":[92],"avoid":[94],"overflow,":[96],"leading":[97],"non-negligible":[99],"overhead;":[101],"2)":[102,152],"training":[105,115],"weight":[110],"outliers":[111],"make":[113],"quantization-aware":[114],"(QAT)":[116],"difficult,":[117],"resulting":[118],"in":[119],"inferior":[120],"model":[121,169],"accuracy.":[122],"To":[123,174],"address":[124],"these":[125],"challenges,":[126],"co-optimize":[128],"both":[129],"At":[134,153],"level,":[137,156],"series":[141],"of":[142],"graph-level":[143],"optimizations":[144],"for":[145],"minimize":[149],"communication.":[151],"develop":[158],"mixed-precision":[160],"QAT":[161],"layer":[165],"sensitivity":[166],"optimize":[168],"accuracy":[170,213],"given":[171],"constraints.":[173],"accommodate":[175],"outliers,":[177],"further":[179],"2PC-friendly":[182],"re-weighting":[184],"increase":[187],"representation":[189],"range":[190],"without":[191],"explicitly":[192],"increasing":[193],"widths.":[195],"With":[196],"experiments,":[198],"demonstrates":[200],"11.7x,":[201],"3.6x,":[202],"6.3x":[204],"reduction":[206],"1.29%,":[208],"1.16%,":[209],"1.29%":[211],"higher":[212],"compared":[214],"state-of-the-art":[216],"frameworks":[217],"SiRNN,":[218],"COINN,":[219],"CoPriv,":[221],"respectively.":[222]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-02-26T00:00:00"}