Life, the university and everything

A Google-funded research expedition is allowing CMU faculty and students to live inside the ‘Internet of Things’—and understand the promise and perils of unprecedented interconnectedness

Ph.d. student Adrian deFreitas (CS 2015) and HCII Director Anind Dey are part of a team of CMU researchers leading an initiative to create a robust platform that will enable Internet-connected sensors, gadgets and buildings to communicate with each other.

If you woke up this morning and asked your Amazon Echo for a weather forecast or relied on your Nest thermostat to get your house toasty, you’re already living in the world of the “Internet of Things,” or IoT, in a way that would have been unimaginable as recently as the start of this decade.

As the Internet of Things swiftly marches from concept to reality, Carnegie Mellon scientists are leading a Google-funded expedition that will use the campus and the surrounding city as a test bed to build new uses for the technology while protecting users’ privacy and safety. There are four broad research areas being explored—machine learning, end-user applications, use of the campus as a “living laboratory,” and the privacy and security implications of the technology.

For the first year of research, CMU received $500,000 from Google to be the lead university in the project, code-named “GIoTTo,” working alongside Cornell University and the University of Illinois at Urbana-Champaign.

“The fact that [Google] wanted to build an open framework was really compelling to us,” says project leader Anind Dey, director of CMU’s Human-Computer Interaction Institute. “Any one company that wants to own all of IoT isn’t going to succeed, and Google recognizes that nobody’s going to own this and you’re better off having an open platform that anyone can connect to.”

Google’s goals for that open platform include ubiquitous inexpensive sensors, better user interfaces for those sensors, and the creation of tools to help end users come up with their own uses for all of that new technology. At CMU and its partner schools, those goals are spread across a variety of teams focusing on specific tasks, including one very big one: maintaining privacy and security in a world where it’s possible to collect so much data on any individual’s behavior.

“We treat privacy as a first-order concern,” Dey says. One way to keep sensitive data under control is the specific project of Mahadev Satyanarayanan, better known as “Satya,” CMU’s Carnegie Group Professor of Computer Science. Satya is focusing on developing the concept of “cloudlets,” miniature data centers that could live in individual homes or offices to manage the data that’s passed from local sensors to a broader network.

“If all of your data leaves your home and goes straight to the cloud, you have no opportunity to control the exposure of that data,” he says. “It’s quite easy, for instance, to find an IoT sensor that you can attach to your water meter and every time water is used, it’s able to report back that this much water was used. Why is it anyone’s business how many times I flush the toilet each day?”

Instead, he says, a cloudlet could manage privacy policies locally to determine who gets access to which categories of data. In addition to controlling privacy, a cloudlet can also reduce the bandwidth demands that the rise of IoT could put on the entire internet.

“Motion sensors, water sensors—the amount of data they generate is small, maybe a few tens of kilobytes every few minutes or seconds. But some of these sensors are video cameras, and if I have 10 cameras in my house or maybe 200 cameras that are watching, you’re talking about pretty substantial data rates,” Satya says. “Transmitting that to the cloud is not economically feasible.”

Instead, a cloudlet could collect all that data and then use locally-determined rules to focus only on what its owners find important.

“Suppose you had face-recognition capability running on your cloudlet at home,” he says. “If the faces it recognizes are your family members, you’d say, don’t bother shipping their faces to the cloud. But any stranger on your property, that’s worth shipping”—at which point other pieces of the IoT landscape could alert a security company or a user’s mobile device.

Making those facial-recognition capabilities work, as with so many pieces of the emerging IoT ecosystem, requires the sort of test bed for which a curious campus such as CMU is ideal.

“We have people that are technical and people that are non-tech, so that provides us with a really nice mix,” says Dey.

In its initial phases, CMU’s expedition leaders are exploring some of the problems that are endemic to campus life.

“For this project, we have seven faculty members that come together for a meeting,” Dey says. “On any given week I have no idea who’s in town and who isn’t. So the question I always have is, do I have time to get a cup of coffee before everyone gets there?”

Using a combination of Amazon’s Echo devices as well as other sensors, the GIoTTO team has developed an application that figures out where other meeting participants are, whether or not they’re planning to attend the meeting, when they’ll arrive—and, most critically, how long the line is at the cafeteria where he might or might not be able to sneak in that cup of java before they converge at his office. Along a similar line of inquiry, Satya imagines a situation in which he’s in his own office and needs help, say, moving a heavy package. Instead of sending an e-mail or text that would go to an entire office’s worth of people, IoT sensors could suppress the sending of that message to anyone who’s not actually in the building and able to help.

Projects like that might sound trivial at first, but Dey says they’re actually designed to test some bigger ideas. For instance, how many sensors are really needed to get useful information about a home or office? Instead of needing a sensor to determine whether his window is open or closed, for instance, the GIoTTO team is creating applications that could use existing sensors—ones that might be measuring air pressure or temperature— that could be trained to detect the subtle differences between a room with an open window and one where the window is closed.

Dey and his team are already working on the ways in which all of this technology could eventually reach consumers. In mid-November, a CMU team traveled to Google to present a first-generation version of an “IoT app store” that could tie together many of their concepts in a way that average users could easily tap. In the months to come, Dey says, the team will be expanding its reach to test more of its concepts in more of the real world.

“We’re really excited to deploy this to a large part of the CMU campus,” he says, “and then to Pittsburgh to help make this a smart city.”

For More Information: 

Jason Togyer | 412-268-8721 | jt3y@cs.cmu.edu