Hybrid Quantum Graph Neural Networks for Robust Botnet Detection in Modern IoT Ecosystems

The proliferation of IoT devices has facilitated sophisticated peer-to-peer (P2P) botnets capable of evading conventional detection. While Graph Neural Networks (GNNs) provide a powerful relational bias for anomaly detection, they often face representational limitations and over-smoothing in large-scale topologies. This paper proposes Direct Quantum Topological Embedding (DQTE), a hybrid framework integrating topological analysis with the high-dimensional feature space of Variational Quantum Circuits (VQCs). By decoupling graph aggregation from the quantum loop via a pre-computed sparse matrix approach, DQTE circumvents the scalability bottlenecks typical of quantum graph learning. The architecture employs a residual hybrid design to mitigate barren plateaus and ensure stable convergence. Validated on the CIC-IoT-2023 dataset via KNN construction, the model demonstrates superior performance, achieving over 95% balanced accuracy and significantly reducing false negatives compared to classical baselines. These results establish quantum-enhanced deep learning as a robust paradigm for next-generation IoT security.