How vulnerable is battlefield artificial intelligence (AI) to cyber and electronic warfare (EW) attack?
ARLINGTON, Va. – U.S. military researchers are asking industry to find ways of assessing the vulnerabilities of military artificial intelligence (AI) programs to enemy cyber attack.
Officials of the U.S. Defense Advanced Research Projects Agency (DARPA) in Arlington, Va., issued a broad agency announcement (HR001125S0009) earlier this month for the Securing Artificial Intelligence for Battlefield Effective Robustness (SABER) project.
AI security risks
Today there are no ways to assesses deployed military AI-enabled systems for their vulnerabilities to cyber attack, DARPA officials warn; the security risks of AI-enabled battlefield systems remain unknown.
To rectify this, the DARPA SABER project seeks to build an AI research group equipped with the necessary counter-AI techniques, tools, and technical competency to assess AI-enabled battlefield systems.
AI technology has reached a level of maturity sufficient to integrate the technology into U.S. military systems. AI could give battlefield advantage by helping improve the speed, quality, and accuracy of decision-making while enabling machine autonomy and automation.
Yet AI has been shown a vulnerability to an adversary's taking control of its data input, which can lead to data poisoning, physically constrained adversarial patches for evasion, and model stealing attacks.