28 June 2021
The iProov Science function is made up of multiple teams, with each team playing a key role in iProov technology innovation. This month we caught up with Jim Bremner, Head of AI. Jim talks to us about his journey at iProov, from Research Scientist to Head of AI, and shares his insights into what it’s like to tackle the increasing threat of deepfakes.
Hi Jim! Can you tell us about your journey to iProov?
I studied Physics at Imperial College London because I thought physics was really cool. I liked that I could work on something technical that had so many unusual and interesting aspects to it. I graduated with a Physics MSci and like a lot of physics graduates, I enjoyed the studying but felt a yearning for something more applied. My degree supplied all the theory I needed to start my journey with artificial intelligence and machine learning. So I started working on small projects to get me up to date with where the field was at the time. I was then involved in a Machine Learning Fellowship where I was able to nail down the practical aspects of applying my theory to real-life projects. I learned a lot about computer vision problems, involving interesting ways of visualizing and making sense of image data. Which was how I became interested in iProov and face biometrics…
So why did you choose iProov, specifically?
iProov face verification is solving a problem that is really interesting – how does one tell whether someone on the other side of a smartphone isn’t an attacker trying to steal someone’s identity? How can you tell they are real? A solution wasn’t obvious to me at first, so that gave me the sense that there was actually quite a lot of open ground – as opposed to an established industry where the techniques weren’t as disruptive. I knew iProov and biometrics had a great balance of these qualities. iProov also has a really big reach, so I knew that I would be making a difference across a lot of people’s lives.
How has the battle against deepfakes changed since you started working here?
The battle against deepfakes has been my main focus through my time at iProov. Yet when I started at iProov, ‘deepfakes’ wasn’t a commonly known term. So the idea that I’d be working on trying to protect against them was maybe quite novel because they’d only been around for a couple of years at that point. I remember telling friends and family about my new role and they were interested to discover there was actually a need for it.
Fast forward a couple years and not only is it really easy for anyone – even with the basic understanding of technology – to go and generate their own deepfakes from a mobile app, but the tools themselves have become a lot more powerful and work in an increasingly difficult environment. So nowadays you only need a single image of someone (which you can get from social media) to create a pretty convincing deepfake, whereas that wasn’t really the case when I started in the industry.
So you are now ‘The Deepfake Guru’! Walk us through an average day in the fight against deepfakes.
The Science team at iProov is focused on staying many steps ahead of the attackers. So we have to keep our ear close to the ground in terms of what’s happening out there with new deepfakes or synthetic media creation techniques; reading a lot of literature and academic papers and getting ideas from them on detecting different types of attacks.
Post lunch, there’s time for intensive thinking through the problems we’re trying to solve. A lot of the problems aren’t straightforward (which I suppose is why we’re trying to solve them!) and they require a bit of solitary thought. I actually do my best thinking on my cycle home. It gives me that forced time where I can’t actually do anything else apart from think (and maybe think about where I’m going with my bike as well).
So tell us, what’s the best kept secret about working in the AI team at iProov?
Hmmm, I will schedule this into my cycle-home-think-time and get back to you.
Ok, ok. But can you tell us why we should be so worried about deepfakes?
People should be worried about them, especially in digital identity, because we’re seeing deepfake attacks all the time. They’re at the stage now where the quality is so good that if you’re not working hard to stop them, they’ll quite easily pass as real. This is why we invest so much time and effort into them.
I feel we are slightly turning the corner in terms of fake news, deepfakes and misinformation. I think more and more people are realizing they can’t trust what they read online. I’m hoping that the same happens with visual media. By showing people concrete examples again and again e.g. a deepfakes of Trump declaring war, they understand what is possible with technology then hopefully it becomes clear these are a threat.
I hear there are a few opportunities opening up to join iProov’s cutting-edge AI Team! What three qualities does someone need to work in the AI team?
I’d say firstly it is important for someone to be very inquisitive. We have to make sure that we protect ourselves against a multitude of attack types which requires exploration of all different possibilities to try and break your own system. If you’re not curious about certain situations that might arise, then we will leave ourselves open to certain types of attack. So having a natural curiosity is key.
Creativity is really important, because a lot of the answers to the problems we are trying to solve don’t exist. We need to create those solutions ourselves. That requires a lot of creativity in terms of concepts, but also how you might implement these concepts as well.
A third one would be empathy. Empathy in a different context to how it’s normally used: this would be empathy towards the attackers. We have to put ourselves in the attacker’s shoes and try to think like an attacker. This gives you a good sense of what we need to look out for because in truth there are many ways people can attack and understanding their motive means we can stay ahead.
If you feel inspired by Jim’s iProov journey and are interested in joining the iProov team – check out our current vacancies. Follow us on LinkedIn and Twitter to keep updated with new openings – we are always looking for new talent.
And meet more of the iProov team below!