BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Los_Angeles
X-LIC-LOCATION:America/Los_Angeles
BEGIN:DAYLIGHT
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
TZNAME:PDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20230712T175212Z
LOCATION:3004\, 3rd Floor
DTSTART;TZID=America/Los_Angeles:20230711T141000
DTEND;TZID=America/Los_Angeles:20230711T142500
UID:dac_DAC 2023_sess106_RESEARCH1471@linklings.com
SUMMARY:$\mathbf{C^2PI}$: Crypto-Clear Private Inference
DESCRIPTION:Research Manuscript\n\nYuke Zhang and Dake Chen (University of
  Southern California); Souvik Kundu (Intel); and Haomei Liu, Ruiheng Peng,
  and Peter Beerel (University of Southern California)\n\nPrivate inference
  (PI) has addressed the rising concern over data and model privacy in mach
 ine learning inference as a service, however, at the cost of high complexi
 ty. In this work, we propose to lighten PI by introducing an empirical-bas
 ed privacy evaluation that leverages inference data privacy attacks (IDPAs
 ) to evaluate data privacy. We then present an enhanced IDPA, DINA, for im
 proved privacy evaluation. Finally, we leverage the findings from DINA and
  propose $\mathbf{C^2PI}$, which requires only the initial few layers to b
 e performed with PI. Our experiments show that $\mathbf{C^2PI}$ can speed 
 up existing PI frameworks by more than $3$x.\n\nTopic: AI\n\nKeyword: AI/M
 L Security/Privacy\n\nSession Chair: Benjamin Tan (University of Calgary)
END:VEVENT
END:VCALENDAR
