17 March 2017
Blogs by author: Andy Rowland , Head of Customer Innovation: Energy, Resources and Manufacturing, BT.
From car hacks to connected homes, every aspect of our lives is vulnerable to cyber crime. But AI research and deep learning is working to keep us safe.
Is the world a safer place?
It’s now over 18 months since Charlie Miller and Chris Valasek carried out the infamous remote Jeep hack that led to the recall of 1.4 million vehicles. They remotely killed the power of a Jeep on the highway and disabled the brakes at low speed. Things have since gone quiet, so is it safe to assume that the cars we drive and the homes we increasingly connect are safe? If anything, the threat is greater, with organised criminals now looking at how they can exploit ransomware and even considering shorting stocks after they’ve caused brand damage.
The car manufacturers are doing their best to increase security within their vehicles — controlling remote access, restricting the use of apps, and vetting their suppliers. However, problems arise when things are outside their control.
Not everything can be guarded against.
One of the biggest fears is that an infected vehicle visits a main dealer, passes the malware onto the diagnostic equipment, and then cross-infects hundreds of other vehicles before anyone knows. This would get around the controls on remote access and could be a real safety issue.
For example, many modern cars use powerful motors to turn the steering wheel at speed for automated parking. These systems are only allowed to work under 10 kmph, but what if the malware spoofs the vehicle’s speed, so when you’re doing 70 mph on the motorway, the car thinks you’re only doing 5 mph? The remote parking function could then be invoked by a hacker with disastrous consequences.
The need for advanced warning.
In the world of IT security, it’s now impossible to keep the bad guys out, as often the threat comes from within; somebody accidently clicking the link on a phishing email or a compromised employee. So threat intelligence becomes the way to keep yourself safe.
We’re about to start a proof of concept to test software that carries out deep learning when the vehicle leaves the factory. It baselines what’s normal and then monitors the vehicle’s systems for abnormalities.
Once an abnormality is spotted, it protects the vehicle and alerts our security operating centres. Here, we can then start to check if other vehicles have been infected, whether they all have the same software version and if there is any relevant chatter on the dark web. The idea is to get advanced warning of the outbreak so it can be contained, in the same way health authorities control infectious diseases.
The same practice can be applied to other environments like connected homes. At the moment, criminals are using home devices to form bot nets to launch attacks to bring down services like Twitter and Netflix, but what if they started to deploy ransomware instead. If somebody hacked your thermostat, most of us could reset it, but if they interfered with our smart meter, we could literally be in the dark.
If you’re interested in seeing what innovations will help you find these threats, join us at Innovation 2017. We’ll be demonstrating our threat analytics tool, which combines AI and human domain knowledge to help you find the needles in the haystacks.