The US Department of Defense is seeking $30.3 million over five years for Polygraph+, a program intended to add machine-learning scoring and contactless physiological measurements to federal lie detection. The budget request still requires congressional approval and does not identify the final technologies.
The Defense Counterintelligence and Security Agency would use the system for employee vetting and insider-threat detection. Earlier Pentagon prototypes offer clues: one company measures heart and breathing rates with ordinary cameras, while another tracks signals including head movement, facial skin temperature and pore activity. The effort comes as the department has increased polygraph testing in investigations of alleged press leaks.
The scientific problem is more fundamental than sensor quality. Conventional polygraphs infer deception from pulse, breathing, blood pressure and sweat, but those signals can reflect stress for many reasons. Reviews by Congress's Office of Technology Assessment and the National Research Council found limited or weak evidence for screening employees. Results also vary by examiner, can produce false accusations at large scale and may be manipulated with countermeasures.
Machine learning could combine more measurements, but researchers quoted by MIT Technology Review say it lacks trustworthy ground truth: old polygraph records do not establish whether subjects actually lied. Without a physical signal unique to deception, a more complex score may automate uncertainty rather than resolve it.