Extreme Digital Vetting of Visitors to the U.S. Moves… — ProPublica
▻https://www.propublica.org/article/extreme-digital-vetting-of-visitors-to-the-u-s-moves-forward-under-a-new
The Department of Immigration & Customs Enforcement is taking new steps in its plans for monitoring the social media accounts of applicants and holders of U.S. visas. At a tech industry conference last Thursday in Arlington, Virginia, ICE officials explained to software providers what they are seeking: algorithms that would assess potential threats posed by visa holders in the United States and conduct ongoing social media surveillance of those deemed high risk.
Some analysts argue that gathering social media data is necessary. ICE already has a tool that searches for connections to terrorists, according to Claude Arnold, a former ICE Homeland Security Investigations special agent, now with the security firm Frontier Solutions. But, he said, potential terrorist threats often come from countries, such as Iraq or Syria, that provide little intelligence to U.S. authorities. As a result, in Arnold’s view, social media information is all the more important.
Privacy advocates take a darker view. “ICE is building a dangerously broad tool that could be used to justify excluding, or deporting, almost anyone,” said Alvaro Bedoya, executive director of Georgetown Law’s Center on Privacy & Technology. “They are talking about this as a targeted tool, but the numbers tell a different story.”
Bedoya noted that the program outline originally anticipated that the monitoring would identify 10,000 high-risk visa holders a year. That suggests the pool of people under social media surveillance would be many orders of magnitude larger. (ICE officials did not address this point at the conference.)
Last week, a coalition of academics and technologists warned in a public letter that ICE’s interest in using big data algorithms to assess risk is misguided, given how rare it is for foreign visitors to be involved in terrorist attacks in the U.S. That means there’s little historical data to mine in hopes of using it to design a new algorithm. The letter cited a Cato Institute analysis that found that the likelihood of an American dying in a terrorist attack on U.S. soil in any given year was 1 in 3.6 million in the period between 1975 and 2015.
Cathy O’Neil, one of the signatories to that letter and author of “Weapons of Math Destruction,” told this reporter in August that any algorithm a company proposes would come built-in with some very human calculations. “At the end of the day, someone has to choose a ratio,” she said. “How many innocent false positives are you going to keep out of the country for each false negative?”
Social media surveillance would be difficult to carry out without collecting collateral data on thousands of American citizens in the process, said Rachel Levinson-Waldman, senior counsel to the Brennan Center’s Liberty and National Security Program.
“Generally, with surveillance technologies, they are adopted for national security purposes overseas, but are then brought stateside pretty quickly,” she said, citing practices first honed overseas, such as intercepting cellphone calls. “So once there’s some kind of dragnet surveillance tool or information collection tool in place for one purpose, slippage can happen, and it will expand and expand.”