A new system from IT company Unisys could help authorities flag suspicious people and cargo at border crossings and airports around the world in as little as two seconds, the company says.
The system, dubbed LineSight, is designed to process data like airline ticket information, cargo manifests, and information from organizations like immigration agencies and Interpol. It can help agencies determine whether travelers or shipments should be admitted, detained, or subject to further scrutiny, using machine learning to flag anomalous arrivals or those that match known bad patterns.
“You can establish business rules that say, ‘for this type of risk of this type of level, take the following actions,” says John Kendall, director of border security programs at Unisys. “If this one happens, immediately interdict the passenger.”
LineSight, announced Tuesday, will be able to use data like passenger travel histories to spot suspicious relationships between travelers or shipments, like if a traveler has frequently flown with a companion flagged for other reasons. It can also flag shipments that are unusual for their port of origin—like a shipment of computers from a company that doesn’t typically send them from a certain city.
It can use machine learning to flag other atypical border traffic as it pops up, similar to how cybersecurity software might detect strange behavior from a hacked server. That’s an advantage over current systems, which mostly rely on rules either explicitly coded into software or taught to agents to detect certain known bad behavior.
“The process of trying to maintain those new rules gets completely unmanageable over time,” Kendall says.
Exactly what data gets used will depend on particular agencies’ needs, though the software is designed to easily accept new feeds of information. Unisys, which already provides screening systems to U.S. Customs and Border Protection as well as agencies in Europe and Australia, will offer LineSight to its existing clients; in fact, the company says it has already been able to use some of its government clients’ data to test the new software. Which clients will use the new system remains to be seen, though Kendall says Unisys plans to market it to its existing clients.
Unisys, which has worked with CBP to test biometric systems that scan fingerprints and faces at the border, also offers software that can do facial matching and other analytics on video streams in real time. But that sort of data won’t be directly processed by LineSight, says Kendall. However, if a traveler has been identified as a certain person by such a system, that information could be fed into LineSight’s database.
The system also isn’t designed to pull in every possible type of external data, not yet at least. For instance, it won’t slurp in travelers’ social media posts–a pile of data that DHS is now authorized to collect from immigrants and other visitors to the U.S.—though it’s possible that agencies will use that kind of information in follow-up screening.
Using automated tools to search for suspicious people can have its pitfalls: ProPublica reported in 2016 on software designed to flag criminal defendants likely to commit more crimes, finding the algorithms were not only wrong, they were biased against black defendants. And that’s far from the only example of apparently biased AI.
Privacy advocates have also raised concerns about a string of legal rulings that make constitutional rights particularly tenuous at ports of entry, giving travelers fewer protections against excessive searches. That’s led some to argue for legal limits to predictive policing tools at border crossings, including in one recent paper in the NYU Review of Law and Social Change.
Such algorithms could improve accuracy and efficiency, but they also threaten to “dilute the reasonable suspicion standard and increase unintentional discrimination in a way that existing law is ill-equipped to prevent,” writes Lindsey Barrett, a fellow at the Future of Privacy Forum. “This threat is of particular concern at the United States border.”
But, argues Kendall, LineSight should generate fewer false positives than existing rules-based approaches today, meaning fewer travelers, not more, detained unnecessarily. Depending on how it’s implemented, it could even reduce the impact of personal bias by human agents.
“If you’re a legitimate traveler or a legitimate shipper what you should find is that clearance should be faster than ever before,” he says. “It should make a big difference to the traveler or shipper, because the number of actual risky travelers or shippers is very small: it’s less than one in a thousand.”