BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Transcending Transcend: Revisiting Malware Classification with Con
 formal Evaluation - Federico Barbero\, University of Cambridge
DTSTART:20220118T140000Z
DTEND:20220118T150000Z
UID:TALK166615@talks.cam.ac.uk
CONTACT:Kieron Ivy Turk
DESCRIPTION:Machine learning for malware classification shows encouraging 
 results\, but real deployments suffer from performance degradation as malw
 are authors adapt their techniques to evade detection. This phenomenon\, k
 nown as concept drift\, occurs as new malware examples evolve and become l
 ess and less like the original training examples. One promising method to 
 cope with concept drift is classification with rejection in which examples
  that are likely to be misclassified are instead quarantined until they ca
 n be expertly analyzed. \n\nIn this talk\, I will discuss our IEEE S&P 202
 2 paper which proposes TRANSCENDENT\, a rejection framework built on Trans
 cend\, a recently proposed strategy based on conformal prediction theory. 
 In particular\, I will hold your hand through the formal treatment of Tran
 scend and the newly proposed conformal evaluators\, with their different g
 uarantees and computational properties. TRANSCENDENT outperforms state-of-
 the-art approaches while generalizing across various malware domains and c
 lassifiers. These insights support both old and new empirical findings\, m
 aking Transcend a sound and practical solution for the first time.
LOCATION:Webinar - link on talks.cam page after 12 noon Tuesday
END:VEVENT
END:VCALENDAR