"Vanderbilt Team Leads $6.89 Million DARPA Grant to Train Cyber Agents Against Attacks"
Daniel Balasubramanian, a senior research scientist at Vanderbilt's Institute for Software Integrated Systems, will lead a four-year Defense Advanced Research Projects Agency (DARPA) grant to create realistic network environments for training cyber agents to combat advanced and persistent cyber threats. According to Cybersecurity Ventures, the cost of cybercrime globally could reach $9.5 trillion in 2024, with a single data breach potentially costing millions of dollars. As the sophistication and impact of cyberattacks grow, network operators must be able to understand and defend against them in real time. The new DARPA grant, "Reinforcement Against Malicious Penetration by Adversaries in Realistic Topologies (RAMPART)," aims to create an advanced learning environment in which teams can learn to defend against incoming attacks and search for system vulnerabilities. The approach used to train cyber agents involves model-based representations and deep Reinforcement Learning (RL) to develop autonomous agents capable of defending a network against attacks. This article continues to discuss the RAMPART DARPA grant.
Submitted by grigby1