Real-Time Intrusion Detection at the IoT Edge: A Hybrid Neural and Rule-Based Framework with Adaptive Fusion

@article{1722545,
      author = {Beatriz Santos, Omar Farouk, Nisha Kapoor},
      title = {Real-Time Intrusion Detection at the IoT Edge: A Hybrid Neural and Rule-Based Framework with Adaptive Fusion},
      journal = {Iconic Research And Engineering Journals},
      year = {2024},
      volume = {8},
      number = {6},
      pages = {1338-1343},
      issn = {2456-8880},
      url = {https://www.irejournals.com/formatedpaper/1722545.pdf},
      abstract = {We propose a real-time IoT intrusion detector that fuses a Transformer-LSTM neural engine with a signature rule engine through an adaptive fusion controller. Rather than fixing the balance between the two, the controller learns per-flow weights from recent reliability, trusting rules on familiar traffic and the neural model on novel patterns. On IoT-23, BoT-IoT, and CIC-IDS-2018 the system attains up to 99.4% F1 with false positives as low as 0.6%, all within a real-time latency budget.},
      month = {December},
  }