AI for cyber, you don’t need to know what the threat is, just the network, says Darktrace

There is a shortage of talent, and personalisation in the new mantra. Everyone who works in tech knows this. But it is not just a problem for legitimate business, it is a problem for cyber criminals too, how do they address the staff shortage? You may not have too much sympathy for them, but they can relax, take it easy, for they have a friend in AI for cyber. And that makes them formidable indeed.

“Cyber crime is a perfect market,” says Max Heinemeyer, Director of Threat Hunting, Darktrace.  He explained: “Whatever works, if phishing works, they go for it. If ransomware works, they go for it. If cryptocurrency works, they go for it. Whatever makes the big bucks, they go for it.”

It also turns out that thanks to AI, cyber crime is really good at personalising; so if you are a CEO, then ransomware might work quite well for the cyber criminal, stealing data from your machine might be lucrative too, but putting a bitcoin mining tool on your machine is a wasted opportunity. But if there is a server used only occasionally for development work, and it just sits around collecting dust, processing very little, then installing a bitcoin miner, might be a very good enterprise for the cyber criminal.

And AI can help the recruiting-challenged cyber criminal achieve this.

Cyber security training: Is it lacking in the enterprise?

As part of Information Age’s Cyber Security Month, we are looking at the importance of cyber security training and education in the enterprise

And that means the best tool for tackling the bad guys might be AI for cyber too

Max should know, he used to be an ethical hacker, a penetration tester, and an also ex-member of the Chaos Computer Club – CCC — he knows a thing or two about what cyber criminals are up to, and, he, along with his team of 30 odd people scattered around the world, hunt down cyber threats. But how can they do it? The cyber world is vast, you can count the numbers of experts in Max’s team without having to draw breath. The answer: AI, or AI for cyber. That’s what Darktrace is good at.

Darktrace applies Bayesian theory to cyber security imageDarktrace applies Bayesian theory to cyber security

Darktrace learns normal network behaviour to spot anomalies

The company was founded five years ago, that may seem recent to any reader who left school before the iPhone was launched, but in the world of AI, that is positively ancient history. Max has been with the company for three years, making him a veteran. When he joined he was employee number 120, now there are almost a thousand employees. And the company, which recently completed its latest funding round, this time rustling up $50 million, has a valuation of $1.6 billion. It’s technology has been deployed in over 7,000 networks, R&D is headquarter in Cambridge but the company has offices in over 33 countries, with a dual HQ in the UK and in San Francisco.

The company was founded by three very different types: ex spooks, boffins and the money. Max put it this way: “Ex intelligence people, from GCHQ, and MI5 and other agencies trying to focus on cybercrime, catching hackers, concluded that they needed to change the legacy approach to cyber security. So instead of looking for signatures and rules, which clearly does not scale, while at the same time there was a huge skill gap, they looked for another way. The spooks realised they needed to change something, so they approached Cambridge Mathematicians, working in AI and Machine Learning, and asked ‘can we work together?’ Investors then completed the three way convergence, so that the company could have the sales force and marketing it needed.”

“You can never anticipate tomorrow,” says Max Heinemeyer  from Darktrace 

So how does it work?

“We use what we call the enterprise immune system; a very simple analogy, which underlines how Darktrace works, relates to understanding cells, instead of looking for what bad looks like, it understands networks, it understands the information age. Drax, the company behind the massive power station, was one if its first customers. And working in that way, Darktrace earned its spurs.And it works by spotting a deviation, as soon as a deviation occurs and attacks, entering a network, it can spot the deviation in behaviour. It’s a form of anomaly detection.

Anomaly detection: Machine learning platforms for real-time decision making imageAnomaly detection: Machine learning platforms for real-time decision making

Anomaly detection is bridging the gap between metrics and business processes to provide more efficiency

So, for example, it may know that a certain individual logs in between 7am and 9am every morning, and goes to Facebook and Instagram, and uses HTPS encryption. Darktrace understand all of this; but if a lot of data is then sent by this individual to an usual website, or it starts making connections to weird servers in say Germany, US, China, Russia or Japan, Darktrace will highlight that.

Diversity may be part of Darktrace’s success. Its founder, Poppy Gustafsson and CEO Nicole Eagan, are both female. In Max’s team, the split is 40% women 60% men.

Cyber security best practice: Definition, diversity, training, responsibility and technology

As part of Information Age’s Cyber Security Month, we look at cyber security best practice – everything from defining it to the importance of training

Puppet masters for the internet

“It’s not the kid working in their parents basement anymore, if we think about the most sophisticated human attacks, they try to blend as stealthily as possible. These subtle nation driven attacks are sophisticated, the people behind them know what they are doing. They are like puppet masters for the internet.”

Max says that the “live off the land”, they don’t operate from a server in say Russia, but from the cloud. They stay very low and move around very stealthily.

Building an effective cyber defence against polymorphic malware imageBuilding an effective cyber defence against polymorphic malware

Jan Van Vliet, VP and GM EMEA at Digital Guardian, goes through the best practices to protect against polymorphic malware.

And from this, he makes a prediction. The AI driven tools cannot make judgement calls as they are automated. We believe that AI driven malware will start understanding context.

“The most the advanced attackers, sit there, they listen, they see what you are doing, they understand ‘oh, this person doesn’t talk to finance very often so I will move around to say Amy’s computer. And Malware can learn to do the same. Narrow AI, like Darktrace understands context, what is normal for a given device, or a normal network or normal environment.”

A CTO guide: Cyber security best practice tips

In this guide, five CTOs provide their cyber security best practice tips – to ensure the best protection against cyber attacks and human error

“If you put yourself in the shoes of a malicious nation stage, then their work is very human capital intensive, you need a lot of very skilled hackers, somebody who can use the tools, to understand Windows, Linux, to move around and attack stealthily. But what if the human attackers, can use a piece of malware to understand by itself, say, that a specific target watches a bit of Netflix on a Friday, so they may create a website similar to Netflix, so by understanding what is normal, hackers can scale much more effectively, and hack into say hundreds of organisation

“So AI can solve the skill shortage of cyber criminals and personalise its activity.”

But if the more sophisticated cyber criminals are moving to AI, that leaves companies trying to resist, with little alternative, but to adopt AI – cyber AI, if you will, too.

Avatar photo

Michael Baxter

.Michael Baxter is a tech, economic and investment journalist. He has written four books, including iDisrupted and Living in the age of the jerk. He is the editor of and the host of the ESG...

Related Topics