A TDD Framework for Automated Monitoring in Internet of Things with Machine Learning

Area of Science:

• Computer Science
• Machine Learning
• Internet of Things

Background:

• Robust, fault-tolerant systems are crucial for Internet of Things (IoT) adoption in critical sectors like finance, health, and safety.
• Existing IoT infrastructure collects vast data for smart cities and homes, presenting an opportunity for intrinsic system monitoring.
• Autonomous monitoring of IoT systems can enhance reliability and performance.

Purpose of the Study:

• To propose a Test Driven Development (TDD) approach for automatic module assessment of IoT devices (ESP32, ESP8266).
• To leverage unsupervised Machine Learning (ML) for autonomous monitoring of IoT device status.
• To develop a framework for evaluating IoT systems and providing insights for monitoring frequency.

Main Methods:

• A framework integrating business drivers, non-functional requirements, engineering view, dynamic evaluation, and recommendations was developed.
• The K-Means clustering algorithm was applied to analyze key IoT metrics: free memory, internal temperature, and Wi-Fi signal strength.
• The approach was evaluated using 25 IoT devices with 15 firmware versions, analyzing over 550,000 data points.

Main Results:

• The ML-based TDD approach successfully monitored IoT device status autonomously.
• It identified device constraint violations and provided insights for optimizing monitoring frequency across different firmware versions.
• The study demonstrates the feasibility of using real-world testbed data for ML-driven IoT assessment.

Conclusions:

• This work presents the first TDD approach for IoT module assessment utilizing ML on real testbed data.
• The proposed method enhances the robustness, fault tolerance, and availability of IoT systems.
• Open-source Python scripts for monitoring and ML model training are provided.