NicoElNino/Getty Images

Pentagon ‘years behind’ industry in using AI to protect networks: official

Top challenges include ensuring validity and cleanliness of data, according to a top cyber leader at the Defense Information Systems Agency.

The Defense Department is "years behind" the private sector in using artificial intelligence to bolster its cybersecurity, a top official said Wednesday.

Drew Malloy, technical director of the Cyber Development Directorate of the Defense Information Systems Agency, detailed the challenges DOD has faced in using AI-based security and said the defense agency was exploring ways to further use emerging technologies in cybersecurity. 

"From my perspective, we are years behind where industry is pushing the forefront of things," Malloy said at an event hosted by the Advanced Technology Academic Research Center. (ATARC is part of GovExec, Defense One's parent company.) He said the main challenges in adopting AI for cybersecurity include ensuring the validity and cleanliness of data.

"There's a gigantic effort that needs to take place within the Department of Defense," Malloy said. "You have to look at the whole end-to-end chain of an AI model to know it from a transparency perspective." 

A Government Accountability Office report published in June urged DOD to instruct its components how to approach acquiring AI technologies and solutions. The report said that DOD has "historically struggled to acquire weapon systems software" and noted that "AI acquisitions pose additional challenges." 

DOD officials have previously underscored the importance of prioritizing an ethical AI strategy when developing and fielding autonomous weapons. Last year, the department established the Chief Digital and Artificial Intelligence Office to help accelerate the adoption and use of data, analytics and AI. 

Malloy said that DISA has benefitted by working in recent years with the Joint Artificial Intelligence Center to "inject machine-learning algorithms to pre-process the data" while detecting any anomalies in real time. 

"You have to really have evangelists within an organization that can help drive people towards" AI, Malloy said. "It's never going to be a top-down thing."