J. Kim
Texas A&M Global Cyber Research Institute, Texas, United States
Keywords: zero-trust architecture, micro-segmentation, multi-robot networks, autonomous threat containment, cyber-physical systems security
Autonomous and networked robotic systems are increasingly central to defense, industrial, and critical-infrastructure operations, yet their resource-constrained endpoints and open network architectures make them highly susceptible to lateral movement attacks once a single node is compromised. This work presents the first practical laboratory instantiation of dynamic, behavior-adaptive microsegmentation for a multi-robot network, developed through a collaboration between the Global Cyber Research Institute at Texas A&M University and ColorTokens Inc., a leading commercial micro-segmentation vendor. Using a physical testbed of Raspberry Pi-powered Turtlebot nodes orchestrated through an enterprise Zero-Trust platform, we demonstrate an automated detection-response pipeline that classifies assets by functional role and real-time security status, then autonomously reassigns compromised robots to a quarantine segment enforcing a block-all policy. Experimental validation confirmed sub-cycle isolation with 100 percent blockage of ICMP, TCP, and UDP traffic to quarantined nodes, elimination of command-and-control exfiltration paths, and immediate restoration of connectivity upon remediation, all without disrupting healthy fleet operations. The architecture decouples tag-based classification and API-driven policy orchestration from the underlying intrusion detector, providing a modular pattern extensible to drone swarms, autonomous vehicle fleets, and other contested cyber-physical environments.