How to Build a Cyber Centaur

Connected Life, F-Secure Life, Security & Privacy

You may be wondering, “Cyber Centaur? What on earth is this guy talking about?”

So let me start by explaining what I mean by “Cyber Centaur.” You might have encountered the “man and machine” concept before in various contexts. It often describes an expert who has mastered his domain being connected to the right kind of tools to make him or her even more productive, enhancing their skills. As you know, a centaur is a mythological creature with the upper body of a human and the lower body of a horse… so kind of an enhanced human in a different sense. What I mean by a Cyber Centaur in our cyber security context is a security expert whose capabilities are amplified and enhanced using software solutions, automation, and — perhaps most significantly — artificial intelligence.

But why do we need Cyber Centaurs in the first place? Why human and machine, why not just one or the other? Can experts not do the job alone? Or could the AI just take over and do everything?

Today – and, I would argue for a good while still – AI is no match for humans in creativity. Nor is modern AI really able to apply learnt knowledge across vastly different scenarios, or even learn in a way that effectively generalizes from a limited number of examples. (Think about how many pictures of cats have been used to train the best image recognition systems and compare that with how many you have seen in your life before you knew what a cat was…) There is certainly a lot of very interesting research in the AI domain going on around improved learning and generalization abilities  — for example look at the work by some of our friends at the Curious AI Company, just to mention one example — but there is still a long way to go.

My background is not in cyber security. I joined F-Secure less than a year ago, but for more than two decades, I have been professionally applying machine learning and artificial intelligence in different environments. In my brief time with the company, I have come to greatly appreciate my colleagues who do cyber security research. They truly showcase how a human expert has an uncanny ability to apply what they have learned to totally new situations in very creative ways. It brings out the little geeky kid in me to be on the inside of an outbreak happening and get to watch what happens behind the scenes in a cyber security company! But even our experts can only do so much. They can only sift through so much data. They can only take so much information into account when making a decision. That’s where AI comes in.

AI, as we all know, is tireless. It will always be ready to go through your billions of events or terabytes of data without rest. It embraces thousands of variables and does complex mathematical computations in a way no human ever could. AI never gets bored of doing the same thing over and over. It is always ready to tackle the next challenge or analyse the next bit of data, and it never loses focus.

We often think about any intelligence, including artificial intelligence, in terms of our inherent point of reference – ourselves. But I personally rather think of this as there being many kinds of intelligence or many dimensions of it. (This is an interesting and even philosophical topic, but perhaps something to discuss more in another blog post.) And even though I very much appreciate the drive towards achieving general intelligence in artificial intelligence research, I don’t feel there is really a need to have that truly “thinking machine” before we can solve complex problems much more effectively with the help of AI – and we are doing this already in many situations.

In other words, you could say current AI is very “narrow” in terms of what it does. It’s very focused, but it’s also extraordinarily effective at what it’s been built and trained to do. But that is not a bad thing! This makes an AI an invaluable partner and companion to an expert. By working together, the human and the machine can accomplish so much more together than either alone, building on one another’s strengths. This has been seen in many domains and is very true also in Cyber Security. So we really want our researchers to benefit and become Cyber Centaurs!

What does this mean in practice? Is there a Cyber Centaur toolkit or mod to download? Or do you need to find that person who is an expert in both cyber security AND artificial intelligence? Or is it enough to give a cyber security expert a link to download a package like TensorFlow and set them loose?

In my opinion it’s (sadly) not quite that easy. Unless you really understand the domain, data and the methods, you are almost certain to reach some type of false conclusions or go in the wrong direction. Thus I believe the solution is actually quite humane – co-operation.

You need the right team and the right approach to the problem. Our cyber security experts, and us in our Artificial Intelligence Centre of Excellence, work to jointly solve different types of problems we encounter. We do this by sharing understanding of threats and evaluating what our data scientists discover via algorithmic means, building models and solutions to address the needs and iterating this so that we can find the best solutions together. (Well, we do also have fellows who possess the elusive combination of cyber security and algorithmic experience. But that’s pretty rare. Teamwork also gives us the benefit of merging totally different viewpoints informed by cross-domain expertise, all which help equip us to face down the next cyber threat and keep our customers safe).

We also have the luxury of being able to collect data on the decisions our experts make when faced with intrusions or malware outbreaks. Labelled real world intrusion data is something that many cyber security machine-learning experts will admit is one of the hardest things to obtain. And it’s crucial to have high-quality training data for complex problems. We combine in-depth understanding of why and how algorithms work with the experience needed to decide when and where we should trust underlying assumptions. This robust approach is a great foundation to solve even the most complex problems and detect even the most elusive threat actors.

So, it’s really quite simple… experts of different domains come together in a positive spirit to solve problems together. Human experts becoming even better, with the help of machines, and helping the AI get even better. Cyber Centaurs are built with skill, dedication and most importantly co-operation and fellowship.

Would you like to be a part of building Cyber Centaurs? We are hiring, check out our open positions!

Matti Aksela
Vice President, Artificial Intelligence
F-Secure Corporation


Excellent, comprehensive discussion of an innovative application of AI to security, blended nicely with the human dynamics.

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

You might also like