Intrusion Detection for Edge-IoT Using LSTM-Autoencoder

This work presents an innovative Intrusion Detection System (IDS) for Edge-IoT environments, based on an unsupervised architecture combining LSTM networks and Autoencoders. Deployed on Raspberry Pi 4, our solution achieves an F1-score of 0.96 with 42 ms latency and detects anomalies, including zero-day attacks, with 97.2% accuracy on the TON_IoT and NSL-KDD datasets. Compared to CNN or Random Forest-based approaches, it consumes 40% fewer resources. A comparative analysis with Snort and Bro also reveals superior energy efficiency (1.8 W vs. 3.2 W) and better adaptability to dynamic environments.