How Cyber Security and AI Can Strengthen Each Other?

Artificial intelligence (AI) and cyber security are two technologies that have become increasingly important as the digitalization of society increases. But is it possible to combine the techniques, and what benefits can it offer? We let two of RISE's real-world experts in each area, Shahid Raza and Daniel Gillblad, explain in more detail and explain how artificial intelligence and cyber security can benefit and strengthen each other.

As more and more units in our everyday lives become connected, the need for us to rely on external suppliers also increases. Both cloud service providers, Internet service providers and hardware providers can pose potential security risks.

"It could be a cloud provider that uses data in an unauthorized way or, for example, a smart speaker who listens when it's not meant to do so," says Shahid Raza, director of cybersecurity at RISE.

Faster and more secure with distributed analysis

To detect and prevent attacks of this kind and increase security, new AI technology can help, and in a new, distributed way.

"We sometimes talk about moving AI and machine learning to the edge, that is, the edge of the network," says Daniel Gillblad. In practice, this means that we do not need to send all the data to a central unit for analysis, but can make the analysis much closer to the unit which, for example, is at risk of attack. This makes the analysis faster and more secure because we also do not have to send potentially sensitive data.

"We can then use this distributed technology to solve security problems in an Internet of Things environment," says Shahid Raza.

Easier to identify deviations

Artificial intelligence is also used to know no one and stop intrusion attempts and attacks in traditional IT systems. Where it is difficult for people to predict all conceivable situations that may arise, a security system can instead be trained to learn what normal behavior is and act in exceptional cases.

- AI is a fantastic shortcut when it comes to handling large amounts of data, says Daniel Gillblad. As more and more data-driven systems are developed, security will become even more important than it has been in the past.

- It is especially critical to secure systems where attacks can cause major damage, such as socially critical functions such as electricity and water supply or defense and security, says Shahid Raza. Used to identify disinformation campaigns in social media.

Correct data is needed during training

In the same way that artificial intelligence can create better solutions for cyber security, cybersecurity is central to the development of AI services. When large amounts of data are used to train artificial intelligence, the data must be accurate and not manipulated.

- It's actually relatively easy to deceive AI or machine learning, says Shahid Raza. Very small deviations in the training data are sufficient for the system to ultimately identify an attempted infringement as normal behavior.

Data as decision support places higher demands on security

Ensuring that data is handled safely is nothing new in itself, but has been an important issue for a long time. The difference is that when AI services start to be used as a decision basis for, for example, law, order and security, they must be able to trust that the decisions that are made are based on data coming from a verified source and that the data itself is not incorrect.

- Whether humans or AI lay the foundation for decisions, inaccurate data can, of course, lead to incorrect decisions, but when decisions are made, such as medical treatments or automated vehicles, it must not be wrong, concludes Daniel Gillblad.