Securing the Internet of Things with analytics at the edge


The Internet of Things, that glorious futurescape in which billions of connected devices take much of the work and tedium out of daily living.

As human beings, we’re addicted to our stuff and what it does for us. So a world in which most of our cell phones and other devices are smart enough to make decisions on their own can’t help but be a better one. Right?

IoT3Well, partly. Our ability to use huge amounts of data from all sorts of gadgets will give us the ability to improve many areas of life, from commerce to medicine, and from transportation to government.

But as I wrote in my last blog, the world isn’t all sweetness and light, and plenty of people want to do us harm. Hackers can break into anything with an IP address, and recent news stories show it’s already happening. Wired magazine’s intentional hack of a Jeep Cherokee demonstrated just how easy it is to do.

So if you’re only considering the time-saving benefits of the IoT era, you might also ponder the flipside. What happens when your car gets hijacked? When your fridge or your air conditioning goes rogue? When your insulin pump turns on you?

As the IoT becomes a reality, and moves from industrial to consumer applications that reach deep into our daily lives, the time is right to ask these kinds of questions. Connecting devices just because we can is not a good enough reason to start bestowing them with intelligence. It’s imperative to pause, weigh the benefits against the risks and create security plans up front. It’s our responsibility to think a few moves ahead.

The tools to do so are already there. Thanks to analytics, we can fight all kinds of cybercrime. We can stream, collect and store data in low-cost environments. Advances in highly scalable in-memory analytics allow us to produce insights and predictions in near real time.

This ability to detect variances from the norm allows us to spot and catch the bad guys, from terrorists engaging in bank fraud schemes to someone trying to tamper with your Fitbit.

So that part is set. What I see missing, however, is a greater sense of foresight. Why didn’t the carmakers design their digital systems to be more secure? It’s a shame that we’re thinking of this after the fact.

So I call on all those who are designing for the IoT era to remember the need to future-proof. Because we know the hacker challenge is out there, it’s not acceptable to bring products to market unless they’ve been cyber-proofed. In our rush to achieve the noble goals of convenience and progress, we cannot leave the consumer open to harm or attack.

The way forward is by building in something called “analytics at the edge.” By the edge, we mean on or near the device as opposed to in some central storage location. Doing analysis at the edge makes sense when the flow of data is very fast, very dense or largely unchanging. In those cases, you don’t want to waste bandwidth sending it over the network for analysis, so you place the analytics “at the edge” of the device. Putting analytics on or near the device will help protect them, and us, by allowing us to spot abnormalities faster.

And as human beings, that is definitely our responsibility to do. They’re just devices, after all. Let’s use our superior brains to make them smart enough to know when something dumb is happening.


About Author

Carl Farrell

Former Executive Vice President and Chief Revenue Officer

Carl Farrell formerly led SAS teams across more than 60 countries around the world. He was the chief architect behind the long-term vision and operational strategy to ensure that growth and market potential are achieved.

1 Comment

  1. Exactly right. There will be too much data out at the edge to bring it all back to the cloud, plus a lot of it will be valueless seconds later. We can use SAS Event Stream Processing to filter, aggregate, and analyze at the edge, and only bring back that data of value for a second (e.g., ESP centralized), third (e.g., Visual Analytics or Visual Statistics) and fourth (e.g., Asset Performance Analytics) phase of analytics with a model learning feedback loop back out to the edge.

Leave A Reply

Back to Top