My Journey as a PhD Student: Dr Giovanni Cherubin

The CDT programme has been for me much more than a PhD scholarship: it allowed me to train my personal and technical skills, it offered me the opportunity to network with industry partners and to intern with great companies, and it gave me a sense of community.  Also, the CDT funding gave me enough support to attend several conferences around the world, thanks to which my research could flourish.

My journey as a CDT student begun right after I completed my MSc in Machine Learning at RHUL.  I had always had a passion for Information Security, which until then I had only pursued in my spare time.  For the following four years, the CDT allowed me to work on this full-time, alongside with my main interest, Machine Learning.

The CDT gave me the great privilege to attend workshops and conferences, even when I did not have a paper to present there.  This was invaluable, particularly during the first years: it helped me both to select a research topic and to find interesting problems, but also to meet PhD peers with whom I subsequently started collaborating.

I think one of the most important activities of my CDT years has been the "summer project".  During the summer of the first year, we were asked to work on a research project of our choice, but without any strong indications on the topic: we could choose a problem out of our comfort zone, or even outside our usual area of interest.  This gave me the feeling that we could "fail safely": there was not too much risk involved in failing to produce novel research for that project — which on hindsight is the best feeling to have when carrying out research, particularly at the beginning of one's PhD.

For my summer project, I selected a problem I hadn't been able to fully define at the time: can we use Machine Learning to provably measure the security of a network protocol (e.g., Tor) against traffic analysis attacks? I did realise the same approach could have had further
applications, but I chose the problem of traffic analysis for concreteness. Before I started, I felt this project was risky: I had a rough idea of where to start, but I was not sure this problem had any solutions where I was looking; indeed, the literature until then had been very vague and rarely formal on this problem.  Nevertheless, the somewhat "risk-free" structure of the summer project with the CDT convinced me to try this, and it worked well.  Luckily enough, this work ended up being the main building block of my PhD thesis, and even now, as a postdoc researcher, I still working on some of those ideas (by using similar principles, one can measure the security of several attacks other than traffic analysis).  Had I not have been given the time and freedom to explore this idea, I might have never dedicated myself to it.

The CDT programme also put a lot of emphasis on our training. We were offered classes on presentation skills (e.g., delivering talks and academic writing), but also on technical skills (e.g., penetration testing), and on management and networking.  We were also encouraged to take internships: personally, I did a couple of academic visits (I spent some time at Cornell Tech and then at École Polytechnique in Paris) and an industry-based internship — I had a fantastic experience at HP Labs in Bristol.  All in all, I feel the CDT helped all of us developing the skills required for both academic and industry careers.

Right after my PhD, I went for a postdoc at EPFL in Switzerland, where I currently work at the intersection between Machine Learning and privacy.