Crowdsourced: These screenshots show privacy ratings provided by volunteers for two different apps.

One of the key recommendations in the U.S. Federal Trade Commission’s recent consumer privacy report is that the mobile industry give people simple ways to opt out of tracking, and communicate the often substantial privacy threats associated with many smart-phone apps. But that’s a tall task, given the expanding universe of more than a million apps—many of which can grab personal information such as the user’s location, phone number, and contacts list. Even when given a chance to review these settings in apps, most people don’t bother—they just click their assent, research has shown.

Now researchers at Carnegie Mellon and Rutgers University are finding it is possible to effectively outsource this screening task and then give consumers intuitive warnings about dubious data-access settings. The prototype system, built for Android apps for now, sends the app’s description—and its associated data-access requests—to human workers through Amazon’s Mechanical Turk crowdsourcing service. 

Those remote workers read the settings and offer opinions—a process that takes them less than a minute and for which they are paid 12 cents apiece. Their opinions can then be aggregated into warnings.

Consider the “Brightest Flashlight” app, from Goldenshores Technology. This app merely creates a white screen to provide illumination, but the app’s permission screen requires you to allow the app to furnish your “approximate location” to advertisers. It can also see your precise GPS position, and asks for your phone’s unique identifier number.

When the CMU and Rutgers researchers had 170 Mechanical Turk workers read the settings on Brightest Flashlight, between 90 and 95 percent of them were surprised by these requirements. A prototype warning screen delivers their findings—complete with red flag icons. “The basic idea here is: How do you help people who are not experts in network and computer security understand what an app is doing?” says Jason Hong, a CMU computer scientist who is one of the leaders of the project. “You are outsourcing people to read privacy settings and tell you what is interesting about it.”