The problems of border control mirror the challenges faced by battlefield intel analysts. Both have a fundamental need to scan vast crowds of faces, to peer into containers and vehicles and make snap assessments about likely risk.

Unisys recently unveiled a software product intended to enhance border agents’ ability to make the right call, even when faced with high volumes of information. Military experts at the company say the technology could form the basis for similar battlefield solutions.

“What action should I take next? Who should I alert? That’s a border control problem, but it’s a military problem as well,” said John Kendall, Director of Border and National Security Program, Unisys Global Public Sector.

Unisys’s latest offering, LineSight, uses advanced data analytics and machine learning to deliver improved accuracy, helping agents decide when it’s necessary to make a closer inspection of travelers or cargo shipments. The system should also help to reduce the number of false positives that waste resources and delay clearance of legitimate travelers.

The problem here is one of volume: U.S. Customer and Border Protection says its officers check more than 1 million travelers and $6.5 billion worth of products on an average day.

Yet only one in every 1,000 travelers or shipments represents a legitimate threat.

“The challenge is in trying to find those needles in that haystack,” Kendall said. “You can’t stop and do a detailed investigation of every piece of cargo or every traveler that comes across. So you have to know in advance which ones represent a likely threat.”

The government has been looking for a technological fix to the problem for some time. In late 2016 for example, The Department of Homeland Security’s Science and Technology Directorate announced a $162,302 award to startup Tamr to build an open-source assessment system to help parse air traveler information. At the time, CBP Deputy Commissioner Kevin K. McAleenan said Tamr’s data analytics capabilities could “improve international border security and enhance the international travel experience.”

President Trump himself has taken an interest in the issue, visiting CBP’s National Targeting Center in February and declaring it to be “quite a facility.” Located in Sterling, Virginia., the Center uses data tools to rapidly compare passenger and cargo manifests against databases and other records, to ferret out potential high-risk travelers.

Military implications

Most of this work today relies on pattern matching, an automated process of scouring data in search of telltale signs that something may be amiss.

“You look for a pattern similar to what you have seen in the past,” Kendall said. “We know that drug smugglers tend to buy their air tickets within 24 hours of departure. They may go through an intermediate country rather than going direct. So you look for those things.”

The problem with this is that analysts are always reading yesterday’s news, with searches based on how the bad actors behaved in the past, rather than in the present.

Unisys proposes taking a deeper dive, augmenting pattern matching with data culled from watch lists and other sophisticated, data-driven approaches. “The statistical analysis is basically looking for anomalies, something that does match any known pattern. Say you have a shipment from a port that you would expect, but it’s from a company that you wouldn’t expect,” Kendall said.

In practice, the algorithms could, for example, help curb human trafficking. “Someone brings a niece into the country, they may go unnoticed,” Kendall said. “But if they keep bringing in different nieces, then the software will see that as potentially a human trafficking situation.”

All of this could have military implications.

In its present iteration Unisys has tweaked the algorithm to seek out activities that might be a red flag for things likes drug smuggling and illegal migration: border-control issues. But a different script could easily be applied to gear the data-analysis activity toward a battlefield scenario.

“In a military environment the process would be the same. You have lots of data coming in and you have to sort through it in a very short period of time. So you have to do the known-threat analysis and you start out by defining the parameters: These are the kinds of things we are looking for,” Kendall said.

Real-time, threat-based analysis could help the military to sift intel in real time. Depending on how the algorithm is defined, software might be able to isolate suspicious patterns of movement among crowds or vehicles. Data analysis could be applied to battlefield video feeds or used to analyze drone footage.

At the end of the day, though, such tools can only point the way for human operators. “It’s a triage system,” Kendall said. “It is not making the final call, but rather making the preliminary analysis on what I am supposed to do next. But it can do that very quickly.”