The Latest / Data Security News

AI Security Put to Test by DARPA

July 20, 2016

The U.S Defense Advanced Research Projects Agency (DARPA) is next month putting AI cyber security systems to the test. The agency is hosting the Cyber Grand Challenge in Las Vegas at the DEF CON hacking convention.


Challenge accepted

The challenge pits seven finalists in a 10-hour Cyber-test. They have been asked to create a machine-learning software program that can not only quickly detect security bugs and flaws, but write patches and implement them on the fly.

The agency estimates that, at the moment, hackers exploit security flaws that on average take some 312 days for humans to notice. By that time, the damage can be huge. That’s why the security agency is holding the Cyber Grand Challenge. It’s looking to inspire and push forward research into cyber security that is automated, fast and totally human free.

Baby robotic steps

The agency is offering a prize of $2 million to the winning team and further prizes of $1 million and $750,000 to those in second and third place respectively. Mike Walker, DARPA program manager said it is highly unlikely any team will make a system that can discover all the flaws and bugs in the challenge. He said:

“Is it possible the systems will fail at the start line? Every Grand Challenge we’ve had indicates that the answer is yes. Autonomy is incredibly hard and autonomy for the first time is breathtakingly hard.”

But of course that is the point of the exercise. As Walker recognises, finding a fully working solution is not really the aim of the Cyber Grand Challenge. The real aim is to inspire and encourage further research and investment into such solutions. Just take the 2004 self-driving car challenge in 2004; no entry in the competition that year went very far before crashing. But we all know how far research into such technology has grown and developed and excelled since.

AI security is an interesting technological avenue to explore, especially in cyber security. At Global Data Sentinel we use automated systems to detect uncharacteristic and suspicious behavior. This allows us to quickly shut down access from suspect sources and limit damage or potential data loss.

We look forward to seeing what the finalist bring to the table next month in the Cyber Grand Challenge.