BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:On the Security and Privacy of Collaborative Learning
DTSTART:20220502T160000
DTEND:20220502T180000
DTSTAMP:20260511T044843Z
UID:060082c5ec9214ea295837288539a4cac3fbd4aedb58f92b9038bd09
CATEGORIES:Conferences - Seminars
DESCRIPTION:Mathilde Raynal \nEDIC candidacy exam\nExam president: Prof. 
 Martin Jaggi\nThesis advisor: Prof. Carmela Troncoso\nCo-examiner: Prof. A
 nne-Marie Kermarrec\n\nAbstract\nCollaborative Learning can be used by sep
 arate\nparties to train a model that will benefit from all private local\n
 datasets. Because parties can be corrupted\, a Collaborative\nLearning sys
 tems need to provide guarantees on the security\nof the final model\, i.e.
 \, on what it will converge to\, and on the\nprivacy of the local datasets
 . The three papers we survey highlight\ndifferent attack points of such sy
 stems\, exploit them to implement\nprivacy attacks\, and finally discuss p
 otential defenses. All strongly\ncontribute to the understanding of some o
 f the privacy risks\nassociated with Collaborative Learning\n\nBackground 
 papers\nPrivacy Risk in Machine Learning: Analyzing the Connection to Ove
 rfitting\nUnderstanding Membership Inferences on Well-Generalized Learning
  Models\nExploiting Unintended Feature Leakage in Collaborative Learning\
 n\n 
LOCATION:BC 233 https://plan.epfl.ch/?room==BC%20233
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR