Hardware-assisted Heartbeat Mechanism for Faulty Device Identification in Large-scale IOT Systems
With increased inter-connectivity among heterogeneous devices, such as Internet-of-Things (IoT) devices and other computing systems, there is a need to ensure that any fault in the deployed devices can be detected in (near) real-time. However, automatically detecting device failures in a large-scale, complex IoT system is challenging. In this dissertation, a faulty-device identification method based on lightweight processor-level architectural support is designed . Specifically, a hardware-based monitoring module is integrated within a processor and connected to a separate monitoring program when an examination is required. By analyzing the information collected by the hardware monitoring agent, the monitoring program determines whether the device being monitored is functioning correctly. Findings from the detailed evaluation show that the proposed approach can detect 90.31% of failures with a minimal hardware overhead of approximately 5k gates. This area overhead is reasonable and would amount to 7.69% of the ARM Cortex-M4 – a lightweight IoT processor – that has a total area (excluding optional caches and scratch-pad memory) of 65k gates.
In addition, a support vector machine-based fault identification technique in IoT sensor is proposed. The proposed approach is based on current and power anomaly detection of sensors. Findings from the detailed evaluation demonstrate that the proposed approach can detect sensor faults with an average 98.76% detection accuracy with 0.45% false alarm rate. Finally, a comparative study of the proposed technique is presented with existing sensor fault detection methods to further highlight the advantages of the proposed approach.