How artificial intelligence protects your business

It was a bizarre hack: In 2017, cybercriminals broke into the network of a casino in the United States through the aquarium’s smart sensors. They took data but did not get the money. Why not? Because the casino used security software based on artificial intelligence. The smart software recognized that something was wrong, and warned about security. Artificial intelligence can thus keep companies safe.

Scan data

“The first application of artificial intelligence in cyber security was the detection of cyber attacks in networks,” says Patrick de Graaf. He works as a cybersecurity expert at the research organization TNO. “It started about thirty years ago. Smart detection software scans large data streams that pass through your network and notices if something strange happens. These flows are too large to control manually, but with machine learning algorithms you can look for deviating patterns. Think of a login attempt at a strange time or a printer suddenly sending a large amount of data. ” Or an aquarium that sucks data away.

Combating data breaches

Security companies also use AI to secure email. Mechanical learning engineer Olga Gureenkova works for Smartlockr, which secures outgoing e-mail with AI. “Our mission is to combat data breaches,” she explains. “Research shows that 88% of data breaches are caused by human error. People send sensitive information via email, which they should not have sent because they do not know exactly what information is sensitive. We want to solve that problem. ”

Smartlockr has developed AI software that links you to standard mail programs like Outlook and Gmail. “When you want to send an email, the AI ​​software detects if the information is sensitive,” Gureenkova explains. “It may contain a citizen service number (BSN) or medical information. Based on the settings you choose, the user will receive a notification with the option to send the email with extra security, for example with two-factor or multi-factor authentication.”

Recognize phishing

You can also protect incoming email with AI. Security firm Barracuda has developed an email application for this. “Based on trillions of emails, our AI software has learned how people and businesses communicate with each other,” said security expert Stefan van der Wal of Barracuda. “What domains, IP addresses and languages ​​does a business use? If a malicious person sends you an email that looks like an email from that company, you may not see that it is a fake message. The software sees it right away. “

As a user, you decide what happens if the AI ​​software detects a phishing email. “You may be notified that this email is suspicious. Or you’re letting the system automatically drag the email to the spam folder. The last option is: throw it directly in the trash. I see customers using this option more and more. “

training of AI

How do you train this kind of AI software? “With large amounts of data or data sets,” says researcher De Graaf. “It is not easy to get reliable, representative data sets. There are open data sources that anyone can use. You can generate the training data yourself, or you make an agreement with a user organization and ask if you can use their anonymised data. ” Smartlockr uses open data sources. Barracuda uses open data sources and anonymized threat data from customers.

AI is indispensable

“AI is indispensable for cybersecurity,” says researcher De Graaf. “Cybersecurity is growing fast, this world is technically quite dynamic. For example, the Internet looks different every millisecond. A large corporate network also changes every second: a device is added, a printer shuts down, and a user disappears. In addition, a lot of data goes through it. It is impossible to keep an eye on. It makes a machine. “

Lack of experts

Automation is also important because there is a shortage of experts, says security specialist Van der Wal. “Globally, there is a shortage of more than a million cyber security experts. At the same time, the attacks are getting bigger, broader and harder to spot. For example, recent cyber attacks use machine learning software. We can only predict those kinds of attacks with AI.”

Sensitive information

AI is also indispensable for recognizing complex sensitive information, says mechanical learning engineer Gureenkova. “If someone mentions a medical diagnosis in an email, such as heart failure or mental health issues, it is not sensitive information. But if the same email also contains information about a particular person, then it is sensitive. An AI model can “Recognize this difference, even though the model sees new information that it did not see during the training process. It is sometimes difficult for employees to see such a difference. Traditional technology cannot either.”

Making decisions

In the future, AI will take over more tasks from human cybersecurity specialists, De Graaf expects. “A cyber attack is often exposed to a high time pressure. You need to make quick decisions. What happens? What should we do? You see AI and machine learning slowly appear in the so-called ‘decision support’, ie in that decision-making. ” AI will also play a bigger role in solving problems, De Graaf expects. “We are now working on technology that will enable software to more frequently repair the vulnerability it has discovered.”

Basically in order

AI already helps you protect your business and will play an increasingly important role. But first and foremost, get to grips with the basics of your safety, advises researcher De Graaf. “Update your systems on time, otherwise there will be a vulnerability. Never use a default password. Make sure you have antivirus software. That base is really the most important. So take a look at more advanced solutions for your security.”

Leave a Comment