DHS Evaluates Potential of Machine Learning for Air Transportation Security

The Department of Homeland Security is exploring how artificial intelligence and machine learning capabilities can support the Transportation Security Agency's mission to protect the American air transportation system. 

The DHS Science and Technology Directorate's Transportation Security Laboratory is evaluating AI and machine learning's potential for improving the detection of concealed threats on passengers and their personal property. 

“These new algorithms will have the potential to outperform the existing rule-based algorithms and present significant opportunities for improving DHS missions that involve efficiently detecting threats that may be concealed among day-to-day travelers and commerce, said Barry Smith, TSL division manager.

He explained that current security algorithms are dependent on manual data mining for translating complex threats into code that is understandable by existing screening technologies.  

Subject matter experts said that more effective algorithms can reduce first-stage alarms in bag screening systems, automate the detection of weapons and concealed threats and reduce secondary passenger screenings and delays. 

TSL is currently developing a roadmap to identify the capabilities, sequencing of activities and resource requirements needed to be included in its efforts to evaluate AI and machine learning. 

Smith added that the laboratory aims to identify and eliminate the weaknesses of its machine learning algorithms to ensure it can correctly detect relevant threats.

TSL is also collaborating with TSA and other groups working on machine learning technologies, opening up the potential for a wider homeland security application.

Lee Spanier, TSL's branch manager for spectroscopy, said other agencies like Customs and Border Protection hope to use such algorithms for detecting shipments of illegal drugs and contraband being smuggled into the United States. 

"Therefore, we need to make sure we thoroughly evaluate these algorithms so that other homeland security agencies can benefit from our efforts and effectively incorporate them into their existing safety protocols,” Spanier said.