Efficient IoT Intrusion Detection System Developed for Resource-Constrained Devices

Strategic Insights -

Researchers at Northwestern Polytechnical University have designed a new intrusion detection system called KronNet, which is capable of operating within the stringent resource constraints of Internet of Things (IoT) devices. KronNet leverages a lightweight feed-forward neural network enhanced with Kronecker product operations to detect various types of attacks in real-time. The system has demonstrated exceptional performance on two popular IoT datasets, achieving high accuracies and low false positive rates.

Key Takeaways:

  • KronNet is a lightweight feed-forward neural network enhanced with Kronecker product operations, designed for real-time IoT intrusion detection.
  • The system leverages Gaussian Mixture Model (GMM)-based oversampling and a hybrid loss function to address class imbalance, ensuring robust detection across diverse attack types.
  • KronNet achieves exceptional performance on the CICIoT2023 and BoT-IoT datasets, with accuracies of 99.01% and 99.91%, weighted F1-scores of 99.01% and 99.91%, and low false positive rates of 0.03% and 0.01%, respectively.
  • The model operates with minimal computational overhead, utilizing 5,074 parameters (19.82 KB) for CICIoT2023 and 4,703 parameters (18.37 KB) for BoT-IoT, with inference times of 0.209 ms and 0.208 ms.
  • Post-quantization, memory usage reduces to 4.96 KB and 4.59 KB, with negligible accuracy degradation (0.06% and 0.01% loss).
  • KronNet demonstrates up to 15,829 x lower FLOPS and 12,010 x faster inference compared to state-of-the-art models, making it a highly efficient solution for edge deployment in resource-constrained IoT environments.
  • The team's work advances IoT cybersecurity by delivering a scalable, accurate, and lightweight IDS capable of real-time threat detection.

Statistics:

  • Accuracy on CICIoT2023 dataset: 99.01%
  • Accuracy on BoT-IoT dataset: 99.91%
  • Weighted F1-score on CICIoT2023 dataset: 99.01%
  • Weighted F1-score on BoT-IoT dataset: 99.91%
  • False positive rate on CICIoT2023 dataset: 0.03%
  • False positive rate on BoT-IoT dataset: 0.01%
  • Number of parameters: 5,074 (19.82 KB) for CICIoT2023 and 4,703 (18.37 KB) for BoT-IoT
  • Inference time: 0.209 ms for CICIoT2023 and 0.208 ms for BoT-IoT
  • Post-quantization memory usage: 4.96 KB and 4.59 KB
  • FLOPS reduction: up to 15,829 x
  • Inference speedup: up to 12,010 x

Sources:

  • KronNet a lightweight Kronecker enhanced feed forward neural network for efficient IoT intrusion detection. Scientific Reports, 2025;15(1):20850.