AI May Pose More Questions Than Answers
Artificial intelligence and machine learning techniques could help information and network defenders recognize patterns of potential attackers so their next moves can be proactively blocked. In addition, cyber tools enhanced with these capabilities could provide a much more detailed picture of the cyber battlefield and increase the potential of success in a cyber campaign. This knowledge would complement the kinetic battlefield and could permit war planners to choose the appropriate mix of cyber and kinetic operations.
The invisible nature of cyber operations makes it a powerful weapon. Unlike its kinetic counterpart, victims usually aren’t aware of an attack until they experience the effects. As a result, much of the most important discussions about U.S. cyber activities must take place behind closed doors in a classified environment. This is one of the reasons AFCEA International is hosting the association’s fourth Classified Cyber Forum.
Jim Richberg, one of the forum’s organizers, says the event aims to address several emerging implications of artificial intelligence and machine learning (AI/ML). Among the topics, event presenters will discuss the prognosis for near-, mid- and long-term development and adoption of AI/ML; adversarial machine learning; and the interaction between defensive and offensive cyber use of AI/ML.
Richberg, who has more than 30 years of experience in the U.S. government leading and driving innovation in cyber intelligence, policy and strategy, says, “The long term development of AI/ML and its impact on cybersecurity are worth examining since it's unlikely to play out in a simple or predicable manner.”
John Gilligan, CEO, Center for Internet Security, agrees that AI/ML in adversaries’ hands will pose challenges unlike those seen today. “Using AI/ML, attackers can refine their methods of attack more rapidly by ‘learning’ about the defenses of a target and quickly turning to other methods of attack that might be more successful,” he says.
AI/ML also may affect kinetic warfare strategic planning, Gilligan adds. “There is the potential that AI/ML cyber tools could provide a much more detailed picture of the cyber ‘battlefield’ and increase the potential of success in a cyber campaign. This knowledge would complement that knowledge of the kinetic battlefield to permit war planners to choose the appropriate mix of cyber and kinetic operations,” he explains.
While cyber and kinetic capabilities may complement each other in the battlespace, Richberg points out several other issues still must be addressed regarding how artificial intelligence and machine learning fit into industry planning when it comes to the government and military environment. For example, he points out developers must still determine where cybersecurity ranks when compared to more readily monetized areas for AI/ML investment.
Richberg also asks, “Are we doing things in AI/ML that may make sense in the near term or from a narrow organizational perspective, but that are suboptimal from strategic/whole-of-nation perspective activities? Are we 'eating our seed corn' as a nation and picking 'winners' prematurely instead of hedging our bets? For instance, many early adapters are making their efforts proprietary and academic research is becoming commercialized. [Do] we lack an adequate academic or commercial training pipeline, etc.?”
These and several other topics about how the future of artificial intelligence and machine learning fit into cyberspace will be explored during the AFCEA Classified Cyber Forum, which takes place on June 12 in Chantilly, Virginia. A detailed agenda and registration is available online.
