Using AI to improve security

robberAI has found its way into many walks of life, but one of the more interesting areas is in cyber security.  Fujitsu and the University of Electro-Communications are working on algorithms that help with the security of its systems.

The creation of water-tight security plans have traditionally been the preserve of experts who can call upon many years of experience.  There has been a gradual introduction of game theory into the mix however, but the very nature of this meant it was often limited to local issues rather than city-wide as data processing becomes a limiter.

The researchers believe they may have cracked the nut however, with their algorithm able to plan 20 times faster for a 100 node challenge and a whopping 500 times faster for a 200 node challenge.  Even that’s relatively small scale however, as a city the size of Tokyo would have more in the region of 200,000 nodes, which the algorithm can crack in five minutes, versus several days for the previous approaches.

Smarter security

The project initially has more constrained ambitions however, and was born out of a desire to better understand how security in enclosed spaces such as airports and railways can be improved.  The theory was that if we know the escape routes of a criminal, we can better control crime in an environment.

Suffice to say, doing this in such complex environments can require huge resources, so the team turned to AI to better optimize the resources they did have.

It forms part of an interesting trend in using AI to improve security.  For instance, I wrote earlier this year about a new Norwegian based project that is using facial recognition technology in airports.  The idea is that rather than going through exhaustive security checks, the technology scans the faces of people inside the airport and automatically runs a security check on them.

Or you have the British start-up that’s using machine learning to do a similar thing inside stores.  They’re tapping into the CCTV footage captured by a retailer and using machine learning to identify criminal behaviors before the individual has left the store.

It’s part of a growing movement to make crime fighting smarter and helping law enforcement agencies make sense of the mountains of data at their disposal.  As systems get more intelligent, it seems to be a trend that will only intensify.

Fujitsu plan to roll out their product commercially sometime in 2017, and it will be fascinating to see how this, and other products fare.


Leave a Reply

Your email address will not be published. Required fields are marked *