Harnessing IoT for Real-Time Anomaly Detection in sUAS Security
Amar A. Rasheed, Mohamed Baza, Gautam Srivastava, Sherif Abdelfattah · 2025
The rapid proliferation of Small Unmanned Aircraft Systems (sUAS) and RFID technologies introduces new threats to national security. While sUAS revolutionizes business operations, it also becomes a tool for malicious actors due to its wireless vulnerabilities. Many Intrusion Detection Systems (IDS) have been proposed to address this, but they often require offline training with heavy processing, making them unsuitable for dynamic mission changes. This paper presents a novel framework that leverages IoT infrastructure (e.g., smart cities) to support rapid, in-field adaptive model training and parameter estimation for IDS-equipped sUAS. To handle the computational demands, we propose a cluster-oriented, distributed training algorithm using LSTM with mini-batch gradient descent, allowing hundreds of IoT devices to collaboratively perform model parameter estimations. The model was implemented on an IoT platform (NXP-Kinetis K64-120 MHz), achieving high prediction accuracy with minimal power consumption and training time, even with contaminated datasets.