{"id":"https://openalex.org/W4416429631","doi":"https://doi.org/10.1109/iccad66269.2025.11240733","title":"Network and Compiler Optimizations for Efficient Linear Algebra Kernels in Private Transformer Inference (Invited Paper)","display_name":"Network and Compiler Optimizations for Efficient Linear Algebra Kernels in Private Transformer Inference (Invited Paper)","publication_year":2025,"publication_date":"2025-10-26","ids":{"openalex":"https://openalex.org/W4416429631","doi":"https://doi.org/10.1109/iccad66269.2025.11240733"},"language":null,"primary_location":{"id":"doi:10.1109/iccad66269.2025.11240733","is_oa":false,"landing_page_url":"https://doi.org/10.1109/iccad66269.2025.11240733","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE/ACM International Conference On Computer Aided Design (ICCAD)","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5032975353","display_name":"Karthik Garimella","orcid":"https://orcid.org/0000-0002-1914-4907"},"institutions":[{"id":"https://openalex.org/I57206974","display_name":"New York University","ror":"https://ror.org/0190ak572","country_code":"US","type":"education","lineage":["https://openalex.org/I57206974"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Karthik Garimella","raw_affiliation_strings":["New York University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"New York University","institution_ids":["https://openalex.org/I57206974"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5030772843","display_name":"Negar Neda","orcid":"https://orcid.org/0009-0008-5701-408X"},"institutions":[{"id":"https://openalex.org/I57206974","display_name":"New York University","ror":"https://ror.org/0190ak572","country_code":"US","type":"education","lineage":["https://openalex.org/I57206974"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Negar Neda","raw_affiliation_strings":["New York University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"New York University","institution_ids":["https://openalex.org/I57206974"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5075674636","display_name":"Austin Ebel","orcid":"https://orcid.org/0000-0003-1150-9219"},"institutions":[{"id":"https://openalex.org/I57206974","display_name":"New York University","ror":"https://ror.org/0190ak572","country_code":"US","type":"education","lineage":["https://openalex.org/I57206974"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Austin Ebel","raw_affiliation_strings":["New York University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"New York University","institution_ids":["https://openalex.org/I57206974"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5087218901","display_name":"Nandan Kumar Jha","orcid":"https://orcid.org/0000-0001-6334-1740"},"institutions":[{"id":"https://openalex.org/I57206974","display_name":"New York University","ror":"https://ror.org/0190ak572","country_code":"US","type":"education","lineage":["https://openalex.org/I57206974"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Nandan Kumar Jha","raw_affiliation_strings":["New York University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"New York University","institution_ids":["https://openalex.org/I57206974"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5089173037","display_name":"Brandon Reagen","orcid":"https://orcid.org/0000-0002-1932-2750"},"institutions":[{"id":"https://openalex.org/I57206974","display_name":"New York University","ror":"https://ror.org/0190ak572","country_code":"US","type":"education","lineage":["https://openalex.org/I57206974"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Brandon Reagen","raw_affiliation_strings":["New York University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"New York University","institution_ids":["https://openalex.org/I57206974"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":5,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.17000743,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"10"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10237","display_name":"Cryptography and Data Security","score":0.9368000030517578,"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":0.9368000030517578,"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":0.027000000700354576,"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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.00989999994635582,"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/homomorphic-encryption","display_name":"Homomorphic encryption","score":0.5795000195503235},{"id":"https://openalex.org/keywords/scalability","display_name":"Scalability","score":0.539900004863739},{"id":"https://openalex.org/keywords/cloud-computing","display_name":"Cloud computing","score":0.5376999974250793},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5325999855995178},{"id":"https://openalex.org/keywords/encryption","display_name":"Encryption","score":0.5281999707221985},{"id":"https://openalex.org/keywords/linear-algebra","display_name":"Linear algebra","score":0.4918999969959259},{"id":"https://openalex.org/keywords/cryptosystem","display_name":"Cryptosystem","score":0.47609999775886536},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.3569999933242798}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7671999931335449},{"id":"https://openalex.org/C158338273","wikidata":"https://www.wikidata.org/wiki/Q2154943","display_name":"Homomorphic encryption","level":3,"score":0.5795000195503235},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.539900004863739},{"id":"https://openalex.org/C79974875","wikidata":"https://www.wikidata.org/wiki/Q483639","display_name":"Cloud computing","level":2,"score":0.5376999974250793},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5325999855995178},{"id":"https://openalex.org/C148730421","wikidata":"https://www.wikidata.org/wiki/Q141090","display_name":"Encryption","level":2,"score":0.5281999707221985},{"id":"https://openalex.org/C139352143","wikidata":"https://www.wikidata.org/wiki/Q82571","display_name":"Linear algebra","level":2,"score":0.4918999969959259},{"id":"https://openalex.org/C6295992","wikidata":"https://www.wikidata.org/wiki/Q976521","display_name":"Cryptosystem","level":3,"score":0.47609999775886536},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.4343000054359436},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.3569999933242798},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.3465999960899353},{"id":"https://openalex.org/C169590947","wikidata":"https://www.wikidata.org/wiki/Q47506","display_name":"Compiler","level":2,"score":0.3325999975204468},{"id":"https://openalex.org/C26713055","wikidata":"https://www.wikidata.org/wiki/Q245962","display_name":"Implementation","level":2,"score":0.31940001249313354},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.29260000586509705},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.2881999909877777},{"id":"https://openalex.org/C519991488","wikidata":"https://www.wikidata.org/wiki/Q28865","display_name":"Python (programming language)","level":2,"score":0.27810001373291016},{"id":"https://openalex.org/C311688","wikidata":"https://www.wikidata.org/wiki/Q2393193","display_name":"Time complexity","level":2,"score":0.27219998836517334},{"id":"https://openalex.org/C153083717","wikidata":"https://www.wikidata.org/wiki/Q6535263","display_name":"Leverage (statistics)","level":2,"score":0.27090001106262207},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.2703000009059906},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.26989999413490295},{"id":"https://openalex.org/C41045048","wikidata":"https://www.wikidata.org/wiki/Q202843","display_name":"Linear programming","level":2,"score":0.2676999866962433},{"id":"https://openalex.org/C118930307","wikidata":"https://www.wikidata.org/wiki/Q600590","display_name":"Tuple","level":2,"score":0.2531999945640564}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/iccad66269.2025.11240733","is_oa":false,"landing_page_url":"https://doi.org/10.1109/iccad66269.2025.11240733","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE/ACM International Conference On Computer Aided Design (ICCAD)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":24,"referenced_works":["https://openalex.org/W236632755","https://openalex.org/W2002555321","https://openalex.org/W2031533839","https://openalex.org/W2134633067","https://openalex.org/W2157999631","https://openalex.org/W2794927157","https://openalex.org/W2889746123","https://openalex.org/W2897925395","https://openalex.org/W2963037989","https://openalex.org/W3033405216","https://openalex.org/W4205474236","https://openalex.org/W4281609193","https://openalex.org/W4281792301","https://openalex.org/W4307823778","https://openalex.org/W4307925365","https://openalex.org/W4315706064","https://openalex.org/W4380881143","https://openalex.org/W4381894550","https://openalex.org/W4389518760","https://openalex.org/W4399851001","https://openalex.org/W4400681761","https://openalex.org/W4402484301","https://openalex.org/W4405184816","https://openalex.org/W4408903490"],"related_works":[],"abstract_inverted_index":{"Large":[0],"language":[1],"model":[2],"(LLM)":[3],"based":[4],"services":[5],"are":[6,205],"primarily":[7],"structured":[8],"as":[9,29,68],"client-server":[10],"interactions,":[11],"with":[12,207],"clients":[13],"sending":[14],"queries":[15,31],"directly":[16,57],"to":[17,49,74,115,118,126,146,167,173,235],"cloud":[18,37],"providers":[19],"that":[20,138,157,201],"host":[21],"LLMs.":[22],"This":[23],"approach":[24],"currently":[25],"compromises":[26],"data":[27,51],"privacy":[28,52],"all":[30,243],"must":[32,70],"be":[33,251],"processed":[34],"in":[35,39,96,124,181],"the":[36,40,75,112,182,222,246],"and":[38,98,134,136,195,199,229],"clear.":[41],"Fully":[42],"Homomorphic":[43],"Encryption":[44],"(FHE)":[45],"is":[46,66],"a":[47,178,189],"solution":[48],"this":[50,83],"issue":[53],"by":[54,80,144,165],"enabling":[55],"computations":[56],"upon":[58],"encrypted":[59,63,196],"queries.":[60],"However,":[61],"running":[62],"transformer":[64,94,239],"inference":[65,95],"challenging":[67],"programmers":[69],"map":[71],"standard":[72],"kernels":[73,91,123],"constrained":[76],"instruction":[77],"set":[78],"provided":[79],"FHE.":[81],"In":[82],"work,":[84],"we":[85,170,187],"explore":[86],"implementations":[87,204],"of":[88,161,192,216,231],"linear":[89,121,129,197],"algebra":[90,122],"needed":[92],"for":[93,224],"FHE":[97,106,116,159,193],"understand":[99],"how":[100],"network":[101],"optimization":[102],"can":[103,250],"help":[104],"mitigate":[105],"costs":[107],"while":[108],"remaining":[109],"performant.We":[110],"leverage":[111],"Orion":[113,172,247],"PyTorch":[114],"framework":[117,248],"benchmark":[119],"several":[120],"order":[125],"profile":[127],"two":[128],"transformation":[130],"methods,":[131],"packed":[132,141],"row":[133,142],"BSGS,":[135],"find":[137,200],"BSGS":[139],"outperforms":[140],"methods":[143],"up":[145,166],"13.7\u00d7":[147],"at":[148],"transformer-level":[149],"scales.":[150],"We":[151,241],"also":[152],"incorporate":[153],"network-level":[154],"pruning":[155],"strategies":[156],"reduce":[158],"runtimes":[160],"feed":[162],"forward":[163],"layers":[164],"11.46\u00d7.":[168],"Furthermore,":[169],"extend":[171],"include":[174],"ciphertext-ciphertext":[175],"matrix-matrix":[176],"products,":[177],"key":[179],"component":[180],"self-attention":[183],"blocks.":[184],"And":[185],"finally,":[186],"perform":[188],"roofline":[190],"analysis":[191],"primitives":[194,208],"transformations":[198],"(SIMD":[202],"encoded)":[203],"memory-bound":[206],"having":[209],"roughly":[210],"0.1":[211],"integer":[212],"operations":[213],"per":[214],"byte":[215],"DRAM":[217],"traffic.":[218],"These":[219],"findings":[220],"illustrate":[221],"need":[223],"exploring":[225],"alternative":[226],"encoding":[227],"schemes":[228],"models":[230],"computation":[232],"within":[233],"CKKS":[234],"unlock":[236],"scalable":[237],"private":[238],"inference.":[240],"conduct":[242],"experiments":[244],"using":[245],"which":[249],"found":[252],"at:":[253],"https://github.com/baahl-nyu/orion.":[254]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2025-11-20T00:00:00"}