BEGIN:VCALENDAR
PRODID:-//DiS Inc.//Colloquium 2.0//EN
VERSION:2.0
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-WR-CALNAME:Colloquia of Digital Security
X-WR-TIMEZONE:Europe/Amsterdam
BEGIN:VTIMEZONE
TZID:Europe/Amsterdam
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
DTSTART:19810329T020000
TZNAME:CEST
TZOFFSETTO:+0200
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
DTSTART:19961027T030000
TZNAME:CET
TZOFFSETTO:+0100
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
UID:670@ds.cs.ru.nl
ORGANIZER;CN=DiS Colloquium:MAILTO:mmarhuenda@cs.ru.nl
DTSTART;TZID=Europe/Amsterdam:20261007T123000
DTEND;TZID=Europe/Amsterdam:20261007T131500
DTSTAMP:20261005T000000
SUMMARY: DiS Lunch by  Andreas Athanasiou
DESCRIPTION: Blending into the crowd: Privacy via Shuffling.  From drawing statistical conclusions to training machine learning models, data collection has undoubtedly become an essential part of modern-day technologies. Over the years, this has led to a rise of legitimate privacy concerns, especially in sensitive applications. In an effort to reassure users, the framework of Differential Privacy (DP) has been widely studied, which offers an upper bound on the amount of information that an analyst can gain about a particular user participating in a system or not. In the recent years, researchers have studied the combination of DP with shuffling, aiming to improve the balance between privacy and utility. The so-called shuffle model relies on the fundamental property that releasing a shuffled vector of the users’ values contains, in general, fewer information than releasing each value separately. In this talk we build upon the blending effect of the shuffle model, enhancing the trade-off between privacy and utility by exploring how this effect can be refined. We examine two settings: location-data, combining shuffling with metric privacy (a variant of DP for domains equipped with a notion of distance) and the distributed machine learning scheme of Federated Learning (FL) for protecting against Source Inference Attacks (which aim to identify exactly which FL client trained a given data point).

LOCATION:the big lecture room in Mercator 1 (MERC1_00.28, ground floor)
STATUS:CONFIRMED
END:VEVENT
BEGIN:VEVENT
UID:669@ds.cs.ru.nl
ORGANIZER;CN=DiS Colloquium:MAILTO:mmarhuenda@cs.ru.nl
DTSTART;TZID=Europe/Amsterdam:20261005T113000
DTEND;TZID=Europe/Amsterdam:20261005T121500
DTSTAMP:20261005T000000
SUMMARY: DiS Lunch by  Markku-Juhani Saarinen
DESCRIPTION: Math ability has consequences -- On AI Cryptanalysis.  Codebreaking with AI: Techniques, work flows, recent results, and reactions from security authorities. A cryptographer breaks down how security assurances of post-quantum cryptography are holding up and which parts might break next.

LOCATION:6.13 at Huygens (HG. 00.075, ground floor)
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR