Artificial Intelligence in Cyber Security: Uses and Challenges

With the ever-growing technological possibilities in artificial intelligence (AI) and the constant threat of hackers exploiting vulnerabilities, we explore the connections between the two and how one can be used as a defence against the other.

For a detailed look at how AI is used in our daily lives, please check out the previous blog article here. In this article, we will discuss how cybersecurity professionals are able to apply techniques of AI in several ways, and the unique challenges that this brings.

The uses and challenges of AI in cyber security are often intertwined. Here are two conflicting use cases.

We are all used to emails asking us if we recently logged in to Netflix, or purchased something on Amazon. This is machine learning checking on us, or identifying a possible threat by using AI. Through AI we can not only recognise the device owner but trigger a warning if there is any chance of a threat. Tracking for self-protection is one thing, but this tracking can also be used by hackers. Just as our behaviour can be tracked, it may also be learned and replicated.

AI-powered biometric recognition (such as voice recognition or recognition of the way we walk) can be a threat as well. Advanced scanning techniques can provide very detailed data on our appearance to third parties. They can also be used for surveillance, profiling and other breaches of our privacy.

Machine learning capabilities and large databases can help to detect threats and vulnerabilities way more efficiently than ever.

Pattern recognition is one basis for AI, for example, monitoring of patterns in web traffic and if something suspicious is recorded, alerts can be sent to the user similar to the first example. These machine learning capabilities work alongside large databases to help detect threats and vulnerabilities. And with the application of machine learning over time, the extensive database is optimized for more precise pattern identification.

Technology advances mean that every day more and more data is gathered and processed, and while limitations such as GDPR compliancy are enforced, third-party entities have more information on their users than ever before.

Everyone is aware that AI surpasses human monitoring capabilities both in terms of quality and quantity and there is the bonus that human error is eliminated. Add on top that machines are able to operate 24/7 as well as having only a few limits concerning high amounts of data and not forgetting all of this can be done almost instantly.

AI can help manage access and maintain an accurate and detailed record of devices, applications, and users with different access levels. It is not uncommon for smartphone users to unlock their phones with face identification, and many smartphones were able to recognise their owners whilst wearing masks during the covid pandemic. This is because biometric identification systems use AI so they can memorize even the tiniest details of your facial patterns. Ultimately this is likely to replace the need for passwords.

AI is beneficial for protecting the entry points of end-user devices. Traditionally this was the purpose of classic antivirus software. A new malware type had to be defined and added to the capabilities of the software by updating it. With the influence of AI technology, we have protective systems that learn how to recognize behavioural patterns of malware or other suspicious processes. This way it can quickly adapt and constantly build up its virtual muscles and discover new malware types and prevent further malware attacks.

As with most new technologies, implementing AI solutions requires a team of experts and these are hard to find as it is a very popular and extensively funded sector in IT. Also, the cost tends to be high, especially when you add in a little risk factor as some of the technology is likely to still be in an experimental stage.

Of course, virus developers use AI too. And with the capabilities of AI technology, a virus can cause greater damage than ever. Virus software is built to overcome antivirus software, attack its code, and bypass it.

AI can as well be employed for social-engineering attacks. Scammers can use it to mimic human language or produce fake images or videos to trick users into sending confidential data. And of course, these techniques can be used for cyberbullying too.

AI is clearly the next step in our technology journey, and we are at an interesting phase where we see how companies intend to use it to move forward. And this is why it is important to keep a constant discussion about what is possible versus what is acceptable.

We all should carefully assess the implementation of AI and focus on building frameworks and agree on standards to stay in control. We can all provide input with the moral machine for example, a project to help society find consensus on which decision an algorithm shall make when in a situation with a moral dilemma.