As you might have noticed, we launched something pretty cool at this year’s SLUSH conference in Helsinki: Project Blackfin. I’ve personally been looking forward to this launch for some time as the project already has some history behind it. I now thought I’d shine a light on that history to help people understand what this is all about, where the idea comes from, and how we’ve gone about actually implementing the first steps on this journey.
My background is not in cyber security. I am a machine learning guy by trade and training. When I joined F-Secure a little over two and a half years ago, I started off by getting a grasp on how things are done in the domain. I naturally started with getting an understanding of what we at F-Secure have done (having machine learning in production since 2006 – now that is quite impressive!), but also reading a lot on what the competition is doing, research in machine learning in the cyber security field, talking with some industry analysts, etc.
As I was doing this, I got a feeling that everyone was doing fundamentally the same things: finding the same type of solutions to the same problems. There were differences, but nothing really radical in my honest opinion. Given the nature of the problems we are looking to solve, I found myself thinking that there must be better ways to approach the industry-wide challenges – in detection and response, but also cyber security in general. I felt that while I fully agreed (and still do) with our approach of combining human expertise with technology (including AI), we as an industry could be doing so much more if we did things a little smarter.
This led to the concept of a distributed swarm of intelligent agents that cooperate and communicate extensively. The concept first saw light in late 2017, some months after I had joined F-Secure. We sparred on it extensively with our Chief Technology Officer Mika Ståhlberg, but the concept was still missing some pieces and it took several more months for it to really come together.
Then finally, one beautiful, sunny day in the summer of 2018 when I was walking along the channel in Helsinki (yes, we DO get very nice weather in Finland, too, at least sometimes), those final pieces just clicked and the rest is pretty much history. I do admit there was a bit of a flurry of internal activity and discussions with all kinds of stakeholders, individual developers, data scientists, product management, leadership team members, and even our Board of Directors – basically everyone we needed to ensure we are all aligned. This has truly been a cross-unit initiative at F-Secure, something we certainly did not do alone in our Artificial Intelligence Center of Excellence, but everyone was onboard and soon we were underway!
The actual development (as in writing lines of production code) of the distributed intelligence platform started in early 2019. We wanted to make sure we worked toward the vision of having true collaborative swarms of intelligent agents (yes, I know that is a mouthful, but it really is also very descriptive of the concept), but also making sure the work would produce immediate value. The vision for Project Blackfin really is something that I find both obvious and revolutionary – it is very different from the common way of approaching the problem, but in a way that feels almost obvious when you think of it.
I am, to be honest, quite often a bit puzzled by the common trend of looking to replicate the way the human brain works in order to build “superintelligence”. Why on earth would we even try to restrict machine intelligence by forcing it to act like a single human? We know the strengths of computers and they are very different from the strengths of humans. Sure, it is only natural to think that since we are intelligent and this is the way we “work” that must be the way to build intelligence. I have no doubt that this thinking is beneficial when taking first steps and starting to understand how things work, and definitely agree that neural networks can be very useful for some problems (like deep convolutional networks in image recognition). In fact, I worked in the Neural Networks Research Centre of Helsinki University of Technology in the 90s.
But, like moving from trying to attach wings to men to building airplanes, at some point we need to think a bit outside the box to find a better solution. We should not try to force human intelligence and machine intelligence to be like the other, but just admit that there are different ways, different intelligences. We should strive to build machine intelligence in the way it is natural to do so based on the strengths of computers – hugely parallel with intense and effective communication, dynamic and adaptive, fast and focused.
Collaborative swarms of intelligent agents, in my opinion, is the path to the future of machine intelligence. Project Blackfin is this path for us.
And as you know, we have released the first generation of Project Blackfin powered technology in our Rapid Detection and Response solution. We are bringing value to our customers through tailored local anomaly detection models which operate together, but also share the global view amongst agents.
This is also the way we will continue Project Blackfin – constantly taking steps forward toward our vision but also continuously bringing value to our customers and partners. Project Blackfin is not just about doing cool stuff (although we do get to do plenty of that), it is first and foremost about protecting our customers in a way that ensures that we all are protected against not only today’s threats but also those to come in the future.