Anomaly Detection on Resource-Constrained IoT Devices Using an Energy-Efficient Dynamic Artificial Intelligence Architecture Enerji Verimli Dinamik Yapay Zeka Mimarisi Kullanarak Kaynak Kisitli IoT Cihazlari Üzerinde Anomali Tespiti
34th Signal Processing and Communications Applications Conference, SIU 2026, İstanbul, Türkiye, 7 - 10 Temmuz 2026, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Doi Numarası: 10.1109/siu71813.2026.11636531
- Basıldığı Şehir: İstanbul
- Basıldığı Ülke: Türkiye
- Anahtar Kelimeler: ADEPOS, Anomaly Detection, Artificial Neural Networks, Edge AI, Energy Efficiency, ESP32, IoT Security
- Recep Tayyip Erdoğan Üniversitesi Adresli: Evet
Özet
Despite their ubiquity, IoT devices often lack robust security. Deploying standard AI security software to edge devices is challenging due to hardware constraints. This study proposes a low-cost solution by adapting the ADEPOS architecture to cybersecurity. Tested on an ESP32, the hybrid architecture uses a lightweight Decision Tree to filter normal traffic and an expert Fast Artificial Neural Network for suspicious data. The system achieved 96.50% accuracy, providing 23.3% net energy savings with a 94-microsecond average processing time compared to the baseline.