{"id":"https://openalex.org/W4293783465","doi":"https://doi.org/10.56553/popets-2022-0109","title":"LLAMA: A Low Latency Math Library for Secure Inference","display_name":"LLAMA: A Low Latency Math Library for Secure Inference","publication_year":2022,"publication_date":"2022-08-31","ids":{"openalex":"https://openalex.org/W4293783465","doi":"https://doi.org/10.56553/popets-2022-0109"},"language":"en","primary_location":{"id":"doi:10.56553/popets-2022-0109","is_oa":true,"landing_page_url":"https://doi.org/10.56553/popets-2022-0109","pdf_url":"https://petsymposium.org/popets/2022/popets-2022-0109.pdf","source":{"id":"https://openalex.org/S4210183172","display_name":"Proceedings on Privacy Enhancing Technologies","issn_l":"2299-0984","issn":["2299-0984"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310320322","host_organization_name":"De Gruyter Open","host_organization_lineage":["https://openalex.org/P4310320322","https://openalex.org/P4310313990"],"host_organization_lineage_names":["De Gruyter Open","De Gruyter"],"type":"journal"},"license":"cc-by-nc-nd","license_id":"https://openalex.org/licenses/cc-by-nc-nd","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings on Privacy Enhancing Technologies","raw_type":"journal-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"hybrid","oa_url":"https://petsymposium.org/popets/2022/popets-2022-0109.pdf","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5102117456","display_name":"Kanav Gupta","orcid":null},"institutions":[{"id":"https://openalex.org/I4210164937","display_name":"Microsoft Research (United Kingdom)","ror":"https://ror.org/05k87vq12","country_code":"GB","type":"company","lineage":["https://openalex.org/I1290206253","https://openalex.org/I4210164937"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Kanav Gupta","raw_affiliation_strings":["Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Research","institution_ids":["https://openalex.org/I4210164937"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5079010262","display_name":"Deepak Kumaraswamy","orcid":"https://orcid.org/0000-0003-0245-9520"},"institutions":[{"id":"https://openalex.org/I4210164937","display_name":"Microsoft Research (United Kingdom)","ror":"https://ror.org/05k87vq12","country_code":"GB","type":"company","lineage":["https://openalex.org/I1290206253","https://openalex.org/I4210164937"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Deepak Kumaraswamy","raw_affiliation_strings":["Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Research","institution_ids":["https://openalex.org/I4210164937"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5077806978","display_name":"Nishanth Chandran","orcid":"https://orcid.org/0000-0002-8528-9768"},"institutions":[{"id":"https://openalex.org/I4210164937","display_name":"Microsoft Research (United Kingdom)","ror":"https://ror.org/05k87vq12","country_code":"GB","type":"company","lineage":["https://openalex.org/I1290206253","https://openalex.org/I4210164937"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Nishanth Chandran","raw_affiliation_strings":["Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Research","institution_ids":["https://openalex.org/I4210164937"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5101907779","display_name":"Divya Gupta","orcid":"https://orcid.org/0000-0002-7288-420X"},"institutions":[{"id":"https://openalex.org/I4210164937","display_name":"Microsoft Research (United Kingdom)","ror":"https://ror.org/05k87vq12","country_code":"GB","type":"company","lineage":["https://openalex.org/I1290206253","https://openalex.org/I4210164937"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Divya Gupta","raw_affiliation_strings":["Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Research","institution_ids":["https://openalex.org/I4210164937"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I4210164937"],"apc_list":null,"apc_paid":null,"fwci":5.1015,"has_fulltext":true,"cited_by_count":48,"citation_normalized_percentile":{"value":0.96008598,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":99,"max":100},"biblio":{"volume":"2022","issue":"4","first_page":"274","last_page":"294"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10237","display_name":"Cryptography and Data Security","score":0.9958999752998352,"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/T10237","display_name":"Cryptography and Data Security","score":0.9958999752998352,"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":0.9940000176429749,"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":0.9918000102043152,"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/computer-science","display_name":"Computer science","score":0.7886671423912048},{"id":"https://openalex.org/keywords/secure-multi-party-computation","display_name":"Secure multi-party computation","score":0.6486347913742065},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5922794938087463},{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.5788857936859131},{"id":"https://openalex.org/keywords/theoretical-computer-science","display_name":"Theoretical computer science","score":0.494639128446579},{"id":"https://openalex.org/keywords/latency","display_name":"Latency (audio)","score":0.4860420823097229},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.3893741965293884},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.3239101767539978},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.23859265446662903},{"id":"https://openalex.org/keywords/computer-network","display_name":"Computer network","score":0.21622949838638306}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7886671423912048},{"id":"https://openalex.org/C18396474","wikidata":"https://www.wikidata.org/wiki/Q2465888","display_name":"Secure multi-party computation","level":3,"score":0.6486347913742065},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5922794938087463},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.5788857936859131},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.494639128446579},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.4860420823097229},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.3893741965293884},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3239101767539978},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.23859265446662903},{"id":"https://openalex.org/C31258907","wikidata":"https://www.wikidata.org/wiki/Q1301371","display_name":"Computer network","level":1,"score":0.21622949838638306},{"id":"https://openalex.org/C76155785","wikidata":"https://www.wikidata.org/wiki/Q418","display_name":"Telecommunications","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.56553/popets-2022-0109","is_oa":true,"landing_page_url":"https://doi.org/10.56553/popets-2022-0109","pdf_url":"https://petsymposium.org/popets/2022/popets-2022-0109.pdf","source":{"id":"https://openalex.org/S4210183172","display_name":"Proceedings on Privacy Enhancing Technologies","issn_l":"2299-0984","issn":["2299-0984"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310320322","host_organization_name":"De Gruyter Open","host_organization_lineage":["https://openalex.org/P4310320322","https://openalex.org/P4310313990"],"host_organization_lineage_names":["De Gruyter Open","De Gruyter"],"type":"journal"},"license":"cc-by-nc-nd","license_id":"https://openalex.org/licenses/cc-by-nc-nd","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings on Privacy Enhancing Technologies","raw_type":"journal-article"}],"best_oa_location":{"id":"doi:10.56553/popets-2022-0109","is_oa":true,"landing_page_url":"https://doi.org/10.56553/popets-2022-0109","pdf_url":"https://petsymposium.org/popets/2022/popets-2022-0109.pdf","source":{"id":"https://openalex.org/S4210183172","display_name":"Proceedings on Privacy Enhancing Technologies","issn_l":"2299-0984","issn":["2299-0984"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310320322","host_organization_name":"De Gruyter Open","host_organization_lineage":["https://openalex.org/P4310320322","https://openalex.org/P4310313990"],"host_organization_lineage_names":["De Gruyter Open","De Gruyter"],"type":"journal"},"license":"cc-by-nc-nd","license_id":"https://openalex.org/licenses/cc-by-nc-nd","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings on Privacy Enhancing Technologies","raw_type":"journal-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4293783465.pdf","grobid_xml":"https://content.openalex.org/works/W4293783465.grobid-xml"},"referenced_works_count":38,"referenced_works":["https://openalex.org/W142258998","https://openalex.org/W182691100","https://openalex.org/W1524288918","https://openalex.org/W2088492763","https://openalex.org/W2108598243","https://openalex.org/W2169004268","https://openalex.org/W2194775991","https://openalex.org/W2212846837","https://openalex.org/W2402235285","https://openalex.org/W2701059868","https://openalex.org/W2765200655","https://openalex.org/W2797583228","https://openalex.org/W2890309926","https://openalex.org/W2902032143","https://openalex.org/W2949140995","https://openalex.org/W2951956578","https://openalex.org/W2953327827","https://openalex.org/W2968502565","https://openalex.org/W2969373235","https://openalex.org/W2974824775","https://openalex.org/W2981672041","https://openalex.org/W2990399857","https://openalex.org/W2990762906","https://openalex.org/W3007848042","https://openalex.org/W3093021001","https://openalex.org/W3094156230","https://openalex.org/W3108672920","https://openalex.org/W3133749795","https://openalex.org/W3151922143","https://openalex.org/W3165750456","https://openalex.org/W3175221797","https://openalex.org/W3177525997","https://openalex.org/W3185154374","https://openalex.org/W3216987697","https://openalex.org/W3217809002","https://openalex.org/W4297952240","https://openalex.org/W4378527334","https://openalex.org/W6836320567"],"related_works":["https://openalex.org/W4388335561","https://openalex.org/W3008625068","https://openalex.org/W4389115113","https://openalex.org/W3128807919","https://openalex.org/W3176411177","https://openalex.org/W2069340834","https://openalex.org/W2550686250","https://openalex.org/W3035501883","https://openalex.org/W2083288298","https://openalex.org/W2607129240"],"abstract_inverted_index":{"Secure":[0],"machine":[1],"learning":[2],"(ML)":[3],"inference":[4,177,233],"can":[5],"provide":[6],"meaningful":[7],"privacy":[8],"guarantees":[9],"to":[10,52,95,110,152,191],"both":[11,111],"the":[12,18,24,41,48,67,73,98,189,194,253,301,321],"client":[13],"(holding":[14,20],"sensitive":[15,21],"input)":[16],"and":[17,83,113,164,168,197,217,264,277,295,307],"server":[19],"weights":[22],"of":[23,47,147,176,193,262,290],"ML":[25],"model)":[26],"while":[27],"realizing":[28],"inferenceas-a-service.":[29],"Although":[30],"many":[31],"specialized":[32],"protocols":[33,115],"exist":[34],"for":[35,159,172],"this":[36,96,198,222],"task,":[37],"including":[38],"those":[39],"in":[40,97,188,330],"preprocessing":[42],"model":[43,101],"(where":[44,102],"a":[45,103,122,149,173],"majority":[46],"overheads":[49],"are":[50,170],"moved":[51],"an":[53,92,228,259,288],"input":[54,106],"independent":[55,107],"offline":[56],"phase),":[57],"they":[58,156,184],"all":[59,186,252],"still":[60],"suffer":[61],"from":[62,201,203],"large":[63,174,315],"online":[64,125],"complexity.":[65],"Specifically,":[66],"protocol":[68],"phase":[69],"that":[70,211,266],"executes":[71],"once":[72],"parties":[74],"know":[75],"their":[76],"inputs,":[77],"has":[78,285],"high":[79],"communication,":[80,293],"round":[81],"complexity,":[82],"latency.":[84],"Function":[85],"Secret":[86],"Sharing":[87],"(FSS)":[88],"based":[89,117,271],"techniques":[90,120],"offer":[91],"attractive":[93],"solution":[94,151],"trusted":[99],"dealer":[100,104],"provides":[105],"correlated":[108],"randomness":[109],"parties),":[112],"2PC":[114,130],"obtained":[116],"on":[118,314,320,327],"these":[119],"have":[121],"very":[123],"lightweight":[124],"phase.":[126],"Unfortunately,":[127],"current":[128],"FSS-based":[129],"works":[131],"(AriaNN,":[132],"PoPETS":[133],"2022;":[134],"Boyle":[135,140],"et":[136,141],"al.":[137,142],"Eurocrypt":[138],"2021;":[139],"TCC":[143],"2019)":[144,306],"fall":[145],"short":[146],"providing":[148],"complete":[150],"secure":[153,232],"inference.":[154],"First,":[155],"lack":[157],"support":[158],"math":[160,248,278],"functions":[161,279],"(e.g.,":[162],"sigmoid,":[163],"reciprocal":[165],"square":[166],"root)":[167],"hence,":[169],"insufficient":[171],"class":[175],"algorithms":[178],"(e.g.":[179],"recurrent":[180],"neural":[181],"networks).":[182],"Second,":[183],"restrict":[185],"values":[187],"computation":[190],"be":[192],"same":[195],"bitwidth":[196,215,219,238,275],"prevents":[199],"them":[200],"benefitting":[202],"efficient":[204],"float-to-fixed":[205],"converters":[206],"such":[207],"as":[208,243,245,318,324,326],"Tensorflow":[209],"Lite":[210],"crucially":[212],"use":[213],"low":[214,237],"representations":[216],"mixed":[218,274],"arithmetic.":[220],"In":[221],"work,":[223],"we":[224],"present":[225],"LLAMA":[226,263,299,335],"\u2013":[227,332],"end-to-end,":[229],"FSS":[230],"based,":[231],"library":[234],"supporting":[235,273],"precise":[236,247],"computations":[239],"(required":[240],"by":[241,311],"converters)":[242],"well":[244,325],"provably":[246],"functions;":[249],"thus,":[250],"overcoming":[251],"drawbacks":[254],"listed":[255],"above.":[256],"We":[257,297],"perform":[258],"extensive":[260],"evaluation":[261],"show":[265],"when":[267],"compared":[268],"with":[269,300],"non-FSS":[270],"libraries":[272],"arithmetic":[276],"(SIRNN,":[280],"IEEE":[281],"S&P":[282],"2021),":[283],"it":[284,313],"at":[286],"least":[287],"order":[289],"magnitude":[291],"lower":[292],"rounds,":[294],"runtimes.":[296],"integrate":[298],"EzPC":[302],"framework":[303],"(IEEE":[304],"EuroS&P":[305],"demonstrate":[308],"its":[309],"robustness":[310],"evaluating":[312],"benchmarks":[316,328],"(such":[317],"ResNet-50":[319],"ImageNet":[322],"dataset)":[323],"considered":[329],"AriaNN":[331],"here":[333],"too":[334],"outperforms":[336],"prior":[337],"work.":[338]},"counts_by_year":[{"year":2026,"cited_by_count":4},{"year":2025,"cited_by_count":20},{"year":2024,"cited_by_count":14},{"year":2023,"cited_by_count":10}],"updated_date":"2026-07-15T18:14:33.161393","created_date":"2025-10-10T00:00:00"}