Nonlinear Coordination-Regime Monitoring for RPAS Human–AI Teams Under Cyberattack

Cyberattacks can disrupt the communication dynamics that support coordination between remotely piloted aircraft systems (RPAS) and ground control stations (GCS), increasing uncertainty for human–AI supervisory control. This study proposes a nonlinear coordination-regime monitoring approach using the RPAS Network Intrusion Detection Dataset. Sliding-window multidimensional recurrence quantification analysis and sample entropy were applied to standardized telemetry features to estimate a continuous coordination-regime state, C t , and its short-term uncertainty. Attack classes produced distinct dynamical signatures: template-like attacks, including Bruteforce, MITM, and DDoS, exhibited more repetitive and structured dynamics, whereas DoS produced a sustained reduction in coordination structure. Window-level detection achieved an accuracy of approximately 0.79, a precision of 0.91, recall of 0.85, and F1 score of 0.88. These findings indicate that nonlinear traffic-dynamics markers can support operator-relevant alert prioritization by distinguishing persistent regime shifts from transient fluctuations. Future work will validate the approach on raw packet streams and evaluate human-facing supervisory decision policies.