Fuzzy-clustering-based multi-scale intrusion detection for the industrial internet of things

The Industrial Internet of Things (IIoT) faces severe intrusion-detection challenges under multi-protocol, multi-device, and massive heterogeneous traffic, especially extreme class imbalance and poor minority-class detection. To address this, we propose a fuzzy-clustering-based multi-scale intrusion detection model (FPBW-IDS). At the data level, we introduce boundary-prioritized FCM-SMOTE: fuzzy memberships distinguish core vs. border samples, interpolation is prioritized in high-membership regions, and sampling weights are adjusted by membership, thereby significantly enhancing minority-class representation and suppressing noise diffusion while keeping the majority class fixed at a multiplier of 1.0. At the feature-learning level, we build three parallel branches across the spatial, temporal, and frequency domains: parallel Pyramid Depthwise Separable Convolutions (PDSC) to capture spatial (cross-feature) dependencies, BiGRU to model temporal dependencies, and wavelet transforms to extract multi-scale spectral statistics; their outputs are fused with learned weighting to form a complementary tri-domain representation. Experiments on the CICIDS2017 dataset show that the proposed method achieves state-of-the-art performance under both unsampled and sampled settings: Accuracy 99.39%, Precision 98.22%, Recall 93.69%, F1 95.56% in the comprehensive comparison; and Accuracy 98.97%, Recall 99.09%, F1 96.69% in a recall-oriented setting, markedly reducing missed detections while maintaining stable training/validation curves and a small generalization gap. These results show the effectiveness and reliability of clustering-guided oversampling combined with tri-domain multi-scale representation for IIoT intrusion detection.