BEGIN:VCALENDAR
VERSION:2.0
PRODID:IEEE vTools.Events//EN
CALSCALE:GREGORIAN
BEGIN:VTIMEZONE
TZID:Asia/Kolkata
BEGIN:STANDARD
DTSTART:19451014T230000
TZOFFSETFROM:+0630
TZOFFSETTO:+0530
TZNAME:IST
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260620T094901Z
UID:B5CE556D-FAD3-4F3C-BF0E-17E7AD02FC8F
DTSTART;TZID=Asia/Kolkata:20260623T120000
DTEND;TZID=Asia/Kolkata:20260623T130000
DESCRIPTION:Abstract: Secure multi-party computation (MPC) holds the promis
 e of enabling privacy-preserving machine learning across data and model si
 los—but in practice\, performance\, scale and useability bottlenecks hav
 e limited real-world adoption. In this talk\, I will discuss how recent ad
 vances in function secret sharing (FSS) are transforming these bottlenecks
  into breakthroughs\, pushing MPC for secure ML from theory to high-perfor
 mance reality. ORCA combines novel FSS-based protocol designs with GPU acc
 eleration to speed up secure training and inference – achieving sub-seco
 nd ImageNet inference. SIGMA brings secure transformer inference into the 
 realm of practicality\, introducing new FSS-based protocols for core ML fu
 nctions and enabling the first secure execution of GPT-class models\, incl
 uding LLaMA2-13B in under a minute. Finally\, I will discuss SHARK\, the f
 irst FSS-based system for actively secure ML inference that outperforms pr
 ior state-of-the-art protocols by two-three orders of magnitude.\n\nSpeake
 r(s): Divya Gupta\, \n\nVirtual: https://events.vtools.ieee.org/m/564489
LOCATION:Virtual: https://events.vtools.ieee.org/m/564489
ORGANIZER:cayantika@gmail.com
SEQUENCE:30
SUMMARY:From Bottlenecks to Breakthroughs: Accelerating MPC for Secure ML
URL;VALUE=URI:https://events.vtools.ieee.org/m/564489
X-ALT-DESC:Description: &lt;br /&gt;&lt;p&gt;&lt;strong&gt;Abstract:&lt;/strong&gt;&amp;nbsp\;Secure mu
 lti-party computation (MPC) holds the promise of enabling privacy-preservi
 ng machine learning across data and model silos&amp;mdash\;but in practice\, p
 erformance\, scale and useability bottlenecks have limited real-world adop
 tion. In this talk\, I will discuss how recent advances in function secret
  sharing (FSS) are transforming these bottlenecks into breakthroughs\, pus
 hing MPC for secure ML from theory to high-performance reality.&amp;nbsp\;&lt;str
 ong&gt;ORCA&lt;/strong&gt;&amp;nbsp\;combines novel FSS-based protocol designs with GPU
  acceleration to speed up secure training and inference &amp;ndash\; achieving
  sub-second ImageNet inference.&amp;nbsp\;&lt;strong&gt;SIGMA&lt;/strong&gt;&amp;nbsp\;brings 
 secure transformer inference into the realm of practicality\, introducing 
 new FSS-based protocols for core ML functions and enabling the first secur
 e execution of GPT-class models\, including LLaMA2-13B in under a minute. 
 Finally\, I will discuss&amp;nbsp\;&lt;strong&gt;SHARK\,&amp;nbsp\;&lt;/strong&gt;the first FS
 S-based system for actively secure ML inference that outperforms prior sta
 te-of-the-art protocols by two-three orders of magnitude.&lt;/p&gt;
END:VEVENT
END:VCALENDAR