September 5, 2014
Topic: Campus News
As use of social media and cloud computing continue to rise, and tools for data mining and facial recognition continue to improve, privacy is at the forefront of both public and academic discourse.
At a symposium to mark the inauguration of President Mark S. Schlissel, leading privacy scholars from U-M and Carnegie Mellon University discussed the complicated concept. The panel was one of two that took place Friday morning.
"Today's symposia are a most appropriate way to recognize President Schlissel's deep commitment to academic inquiry and thoughtful discussion," said U-M Provost Martha Pollack.
Professor Alessandro Acquisti of Carnegie Mellon University was the keynote speaker during a symposium on privacy and identity in a hyperconnected society. (Photo by Scott C. Soderberg, Michigan Photography)
The keynote speaker was Alessandro Acquisti, a professor of information technology at Carnegie Mellon who demonstrated several years ago that an anonymous photo could lead to the subject's Social Security number.
Using geographic and demographic information, his team was able to sift through archives of public photos to connect a name with a face while subjects answered a brief survey. Then, by using the Social Security Administration's list of ID numbers of people who have died, they used statistical techniques to predict the subject's Social Security number.
"If I were an identity thief, I wouldn't use this approach, but as a proof of concept, it is quite powerful," Acquisti said.
In 2010, Facebook users shared 2.5 billion photos per month. Acquisti and his colleagues have shown that these images, even while they seem innocuous to the sharer, could have repercussions. They could illuminate aspects of the user's life that could cloud a potential employer's decision. And they could be shared beyond the people the user intended.
In the United States today, the current approach to privacy revolves around personal control and transparency, but those aren't enough, Acquisti said. Several panelists agreed.
Erin Krupka, an assistant professor at the School of Information, studies social norms and has examined why people take actions that don't seem consistent with their interests. Today's common "public by default, private through effort" arrangement forces users to define and understand privacy themselves. People often take cues from their peers.
"You may have in mind grandma and grandpa looking at your children's photos, but then they share it with others. You may be making reasonable privacy decisions, yet personal privacy is not only in your hands. It's in the community's hands."
Kevin Fu, associate professor of computer science and engineering at the College of Engineering, pointed out that too often privacy and security are afterthoughts for the entities building these systems.
"Google didn't encrypt all its back-end systems until after the Snowden disclosures," he said. "The open source community did not re-engineer OpenSSL until after HeartBleed. It's like insurance. No one wants it until they need it."
Fu called for computer system designers to "bake" these safeguards in from the start.
Regardless of what people share and what security measures are in place, other panelists said, the issue is much bigger.
Privacy itself might be more a problem than it is a solution, says Catharine A. MacKinnon, the Elizabeth A. Long Professor of Law.
In the context of violence against and exploitation of women, for example, privacy keeps domestic abuse hidden and allows for consumption of pornography as a protected right, MacKinnon said.
"Once we understand something like sexual abuse as a violation of gender equality, as pervasive as these privacy concerns are, is it an unworkable form of protection in this sphere and others?" MacKinnon asked.
Indeed, privacy needs to be weighed against other values, said Martha Jones, an Arthur F. Thurnau Professor and associate professor of history and Afroamerican and African studies, and adjunct professor of law.
At times, it could be more important to be included in the data rather than left out.
"I'm thinking about the old-school big data project, the U.S. Census. African Americans are historically undercounted and as a result, they are deprived of political and economic benefits," Jones said. "As we grapple with questions of privacy, how do we do that without losing sight of these competing values?"
The panel was moderated by School of Information Dean Jeffrey MacKie-Mason.