Curriculum
Computer Science for Societal Challenges and Innovation, XL series
Grant sponsor
Fondazione Bruno Kessler; Dipart. di Matematica UNIPD su fondi progetto Locard
Supervisor

Conti Mauro
Co-supervisor
s
to be defined
Contact
umberto.salviati@phd.unipd.it

Project description
Robust AI: AI is being widely adopted in several areas, from healthcare to the automotive industry, and from agriculture to the industrial sector. Many of these applications are sensitive both in terms of safety and security. Therefore, it becomes of paramount importance to understand if and how attackers can exploit such systems and to design more robust ones. This project aims to focus on “adversarial machine learning,” both from the attacker’s point of view, to understand ways an adversary can abuse AI solutions to gain an advantage (with attacks such as model stealing, model poisoning, or membership inference), and from the defense’s point of view, to explore novel techniques to make AI solutions more robust against these attacks.