'Blind' AI is the best solution to solve security problems
Implementing an Artificial Intelligence (AI) system that is 'blind' and has no pre-selected parameters can solve real-life security problems, says Mariana Pereira, Director at Darktrace, an AI specialist company.
Hackers are getting better and systems that rely on past data will not cope anymore.
Many companies still believe that working with an AI system that knows what it is looking for is the best solution to solve security issues but what this does is to limit the way the system works. 'That doesn't work very well in the cyber security,' says Pereira during a talk at the IoT Tech Expo conference in London. Hackers are getting better and systems that rely on past data will not cope anymore.
Darktrace has created an AI that can detect anomalies between devices and stop certain parts of the connected system to repair it. A more proactive AI able to adapt to the changing conditions without looking at a particular area is more useful as it scans the whole system.
'The key is understanding the whole network for all of its parts and have 100% visibility on all systems,' explains Pereira.
'It is often one device that starts speaking a different language.'
The AI cannot expect what is going to go wrong but by spotting weird interactions it is able to notify the users and ask if the interactions should be seen as the norm or as security problem. 'It is often one device that starts speaking a different language,' continues Pereira and gives some example of how the AI solves security issues. A fish tank in a casino in Las Vegas was connected to an Internet of Things network but once installing Darktrace system the company was able to spot that hackers had used the tank to connect to other parts of the company's network and extract data.
AI that goes 'blind' into a company is better able to respond to real problems and help the security team by buying them time.