CYBER-DEFENCE FELLOWSHIPS: Edoardo Debenedetti

• How did you find out about the CYD Fellowships and what motivated you to apply?

I did an internship at the CYD Campus between 2020 and 2021 when I was doing my master’s at EPFL. It was around then that the CYD Campus launched the CYD Fellowship program. At the time, I was already considering pursuing a Ph.D. at a Swiss university. Hence, I decided to keep an eye on the CYD Fellowship program. I indeed ended up doing a PhD at ETH Zürich, and soon after I started my Ph.D., I decided to apply for the fellowship.

• What was your CYD Fellowship project about?

My project is about machine learning security and privacy, under a real-world lens. Systems that make use of machine learning are being deployed to real-world applications with millions of users. These systems already have access to users' data and will soon be able to act on their behalf. During my doctorate, I want to study what the security and privacy risks of these systems are for their users, and how to prevent, or at least mitigate, attacks against them.

• What were the advantages of conducting your doctoral project at the CYD Campus?

One of the biggest advantages is that I get to interact with other CYD Campus researchers, both at the CYD Campus office and at the CYD campus events. Another major advantage is the travel funding: thanks to this, I get to travel to academic conferences even if I am not publishing papers as part of the proceedings. I believe this is a big opportunity to network with other researchers in the field and find out about the latest work.

• Did you as a child dream of working in cyber-defence?

I wouldn’t say since my childhood, but for sure since I was a teenager. I’ve been a bit of a geek since my early teenage years: always messing around with my family’s desktop PC and my mobile phone certainly made me interested in the “cyber” aspect of cyber-defence. As per the “defence” aspect of it, I spent the last three years of high school as a cadet at the military school of the Italian Navy. This contributed to developing an interest in the “defence” aspect of cyber-defence.

• What is driving you to pursue research in cyber-defence?

I believe that cyber-defence is particularly interesting because of its broad range of impact on tech users. At the same time, it’s an ever-evolving field: new technologies must be assessed for their security and privacy features.

• What is the most important lesson you have learned in your scientific career so far?

During my master’s, as part of a semester project, I contributed to a project that, after the end of the semester, we submitted for publication at a conference. The first time we submitted it, it got rejected. People often complain about the peer review process, and how it is random and ineffective. However, that one time, I learned that papers can improved quite a lot between different iterations thanks to reviewers’ feedback.

• What are you most proud of in your career to date?

I believe it’s RobustBench, a project I contributed to during a semester project at EPFL. We worked on the project in 2021, and researchers in my field still use it and find it useful to this date (2 years in machine learning research is a very long time!)

• Outside the lab, what do you enjoy doing most?

I enjoy anything that’s outdoors, from sailing to hiking, climbing, and skiing. When I’m not outside, I like listening to a lot of different types of music and watching movies, both at home and at the cinema.

• What were your expectations about the CYD Fellowships?

I expect the fellowship to help me build solid foundations for my research career, by granting me more independence and opportunities to build a good network.

• Could you share some tips with future applicants who are considering applying for the CYD Fellowships?

Think about something you would be proud to research and/or build during your Ph.D. or master’s thesis; ask yourself what your goals are, and what the potential impact could be. Once all this is clear, writing why it matters and why it is important will be much easier.