Mathematics-driven state monitoring and tampering defense for a hierarchical heterogeneous UAV swarm communication system

State monitoring and data integrity are essential for unmanned aerial vehicle (UAV) swarm communication systems in low-altitude remote auto-operation systems (LA-RAOSs). This study proposes a mathematics-driven hierarchical heterogeneous network (MD-HHNet) framework for state monitoring, fault localization, and timing-based tampering defense. A linearly coupled state generation model represents the propagation and aggregation of node information, while a non-stationarity (NS) measure characterizes changes caused by communication link faults. Observation nodes are selected using predefined fault signatures and localization performance. The top-layer node provides an initial normal/abnormal decision, followed by hierarchical backtracking to localize faults. Under the same model configuration, the evaluated noise-free single-edge faults achieve 100% localization accuracy. At a signal-to-noise ratio (SNR) of 10 dB, the average localization accuracies across candidate observation nodes are 86.46% and 86.11% for the low-layer and middle-layer networks, respectively. When the SNR decreases from 10 to 5 dB, the average MAE changes from 0.36 ± 0.08 % to 0.40 ± 0.16 % for the low-layer subnetwork and from 0.76 ± 0.40 % to 0.81 ± 0.23 % for the middle-layer network, respectively. The localization time measurements have a mean of 47.00 ms and a standard deviation of 18.36 ms. A PID-based tampering model further demonstrates how access to a fault-free value can enable manipulation of faulty outputs. The observed minimum tampering completion time of 120 ms exceeds the maximum localization time of 78 ms, indicating a monitoring advantage under the evaluated implementation. These results provide a topology-level validation of the framework, supported by failure cases analysis and computational and storage complexity analysis.