In the week or so that has passed since Open AI launched its latest research project Chat GPT, examples of how it can be used have been evicted. Users have allowed the tool to write columns, articles or social media posts.
At the same time, there are more and more reports that Chat GPT can become an important tool for cyber security in the future. Machine learning is already a key technology to improve cyber security and at the same time free up time for the own security team.
With this technology, defenders can let machines discover vulnerabilities, discover new attacks, analyze them and quickly roll out protection.
Detects vulnerabilities
After Chat GPT was launched on November 30, there have thus also been many examples of how the service can be used to write code and create applications – but also to discover vulnerabilities in code and applications. Specialists have started testing Chat GPT in security work.
The IT security site Information Security Buzz has collected a series of impressions from the industry like this at the start and where a number of examples are given of how Chat GPT has identified vulnerabilities in applications and even itself came up with suggestions for malicious code that could be exploited.
And when Sans Institute on the technology site Venture beat sharing their top tips on which trends will dominate the cybersecurity market in 2023, they put Chat GPT at the top of the list, arguing that this is “likely to be a game changer.”
Chat GPT simplifies vulnerability management, in terms of flaws caused by poorly maintained network monitoring tools, and it can also help tackle the cyber skills gap.
“Technology will only improve and likely result in fewer vulnerabilities, but also in tools for developers to use to write more efficient code and to productivity,” they state.
Is it all gold and green forests now? No, when Computer Sweden talks to security experts, they emphasize a number of problems and risks with the development.
Can go really wrong
When we ask Jonas Lejon, IT security expert at Triop, about what he thinks Chat GPT will mean for the industry, he says that the tool has the potential to make security work more efficient, but that there are also some problems with this and other AI- tool.
– I have tested it and I would say that it is right in quite a few cases, probably 50 percent. The difficult thing is to judge if and when it is wrong. That’s what I think you can say about these AI tools in general, that of course they help a lot, but you have to have the ability to judge for yourself when it’s wrong.
If that ability is lacking, things can go really wrong. Jonas Lejon continues:
– So I think this can be an aid, but it won’t be the case that it replaces an entire industry. However, it can help an industry, including the security industry. We’ve been looking for tools that can improve efficiency, but so far there hasn’t been much. There you can say that this tool has potential.
Mikko Hyppönen, head of research at Withsecure, formerly F-Secure Business, tells Computer Sweden that he is worried about the development. A lot is happening in the AI world right now, he continues, not only with Chat GPT but also with Deepminds Alphacode.
– I see a couple of advantages for cyber security with this development, but I possibly see even more disadvantages.
Mikko Hyppönen continues:
– The security consultants’ work will be easier. It becomes easier to review code, find vulnerabilities and fix them when you have a helping hand like Chat GPT. Unfortunately, people with bad intentions can do the same thing.
– “Bad guys” can also build, for example, automated CEO scams or romance scams on top of large-scale language models such as Chat GPT. Thankfully, they can’t run their own copy of Chat GPT.
Systems like Alphacode will one day be so good at coding that they won’t make mistakes, according to Mikko Hyppönen.
– That’s good: No more bugs, which means no more vulnerabilities. It also means a lot of unemployed developers and creating malware has suddenly become much easier.
