Who has your data? Researchers scrutinize apps for undisclosed ties to advertisers, analytics companies

If you want to better understand how an app or a service plans to use your personal information, its privacy policy is often a good place to start. But a recent study found there can be a gap between what's described in that privacy policy, and what the app actually collects and shares.

An analysis by University of Toronto researchers found hundreds of Android apps that disclosed the collection of personal information for the app developer's own purposes but, at the same time, didn't disclose the presence of third-party advertising or analytics services that were collecting the personal information, too.

"This is one of the ways in which you're getting tracked through your use of apps," said Lisa Austin, a law professorand one of the study's co-authors.

To generate revenue, app developers often embed software code, known as ad libraries,allowing them to display ads within their app. Because they want to make the ads relevant to individual users, ad libraries often want specific information about those users.

For those who may be more familiar with the cookies that track your online browsing habits, Austin says that on mobile devices,"you're being tracked through these ad libraries and these analytics libraries in a very similar way."

• Why your next job interview could be with a machine
• Researchers say it's time to crack open AI 'black boxes' and look for the biases inside

The researchers have been working on a software project called AppTrans, with the goal of making undisclosed data collection practices more transparent. The software looks for evidence of data collection that isn't spelled out in a privacy policy by comparing the policy'slanguagewith an analysis of the app's code.

It does this, in part, using machine learning artificial intelligence to automatically scour privacy policies for language that points to the collection of location data, contact informationor unique device identifiers. Such data can be useful for targeted advertising or be used to build profiles of user behaviour.

And handing this data to a third party isalso a way for developers to monetize free apps.

Of the 757 apps analyzed, the researchersfound nearly 60 per centof apps collected more information than stated in their privacy policies.

Austin called the finding"eye-opening."

"It was so bad," said David Lie, a computer science and engineering professorand another of the study's co-authors.

The team'sfindings were published in June, and are based partly on work done by one of Lie's graduate students, Peter Yi Ping Sun. The project was funded by the Office of the Privacy Commissioner of Canada.

All or nothing

Under Canada's privacy laws, developers should have to disclose both the information they collect themselvesand information collected by third-party services embedded in the app's code.

"You can't have informed consent if you don't know that your information is being collected by these third parties," saidAustin.

Part of the problem is that while apps typically have to ask the user for permission before accessing sensitive data, such as location or a person's contact list, granting access is all or nothing.

If you give a weather app access to your location for a more accurate forecast, for example, a third-party advertising service embedded in the app could access it, too. It would be up to the app developer to make that possibility clear.

This is technology that we can use to to make the digital world more transparent," she said, "and that's a real win.- Lisa Austin, study co-author

The U.S. Federal Trade Commission has warned app developers that they must clearly explain to users how they plan to share personal information with ad libraries and seek consent before doing soor face potential legal repercussions.

Google similarlyexpectsdevelopersto disclose any data shared with third parties in an app's privacy policy, including marketing partners or service providers as do many third-party libraries themselves.

So why aren't developers doing it in practice? Lie wondered that, too.

The researchers knew it was unlikely that most developers were intentionally lying to their users, he said. Instead, they concluded that app developers are likely just as bad at reading their own privacy policies as their users.

"We can surmise that, similar to how end users often do not read the privacy policies of the applications they use, application developers do not read or properly incorporate the privacy policies of third-party libraries and tools they use to build their applications," they wrote in the study.

Re-imagining the privacy policy

The researchers see automated solutions, such as AppTrans, as one way to help regulatorsbetter scrutinize the deluge of appsbuilt and added to mobile-app stores each year.

They also imagine their project could eventually help developers analyze their own apps for non-compliance.

For users, the hope is that software like AppTrans could shed more light on how their personal information is collected and shared.

Lie says the team isalready at work on a second, more polished version of AppTrans that would recognize when an app is attempting to access sensitive personal information a person's contact list, for example and then display the relevant part of the app's privacy policy that explains the reason why.

He described it as part of a broader effort to reimagine the privacy policy as a more dynamic tool for transparency, rather than a static document that no one reads again after they've installed an app.

For Austin, it's also an example of how artificial intelligence can be used to society's benefit, in spite of very legitimate concerns about algorithmic bias and automated decision-making run amok.

"This is technology that we can use to to make the digital world more transparent," she said."And that's a real win."