The Pentagon's high-tech thinktank DARPA is worried about the future of the Internet of Things. Very worried. In a few years, DARPA believes, Internet-connected cars, home appliances, and medical devices will give criminals and foreign militaries an unprecedented opportunity to wreck havoc. DARPA is trying to train computer networks to be strong enough to defend themselves against hackers—and they want the public’s help to do it.
The agency is launching their Cyber Grand Challenge today, and it has an ambitious goal: to build next-gen technology that works better than current firewalls and anti-virus software in repelling hacker attacks. Competitors in the Grand Challenge are coming primarily from universities and commercial tech companies (though anyone can participate) and winners will be chosen at the 2016 installment of Def Con, a summertime computer hacker convention in Las Vegas that has had a tangled relationship with the U.S. military and government. The top entries in the Grand Challenge will "teach" computer networks how to automatically deflect attacks without needing a human at the controls.
"Right now the circuit [of competitive computer hackers] is dominated by experts, people who are really good at finding deep flaws in software quickly," DARPA program director Mike Walker tells Fast Company. "The champion of this circuit could one day be a machine."
Competitors in the Cyber Grand Challenge will face an uphill race in building smart networks that can ward off danger. This will be a formidable machine learning challenge, simply because of the quantity of data involved.
"Attackers have the concrete and often inexpensive task of finding a single flaw to penetrate a system, while defenders have the hugely expensive task of finding all flaws (thousands?) and patching them before an attack occurs. We believe automation can upend these economics," Walker noted on DARPA's Reddit Ask Me Anything thread.
To enter the Cyber Grand Challenge, go here.
This story was reported with the assistance of Alex Halperin.