Containerized Anomaly Detection Framework for IoT Networks Using Machine Learning

The increasing use of Internet of Things (IoT) devices generates continuous sensor data that requires efficient monitoring and analysis. Detecting abnormal sensor readings is important for identifying unusual behavior and supporting reliable IoT operations. This paper presents a Containerized Anomaly Detection Framework for IoT Networks Using Machine Learning. The proposed system groups continuous sensor readings into fixed-size logical containers and applies machine learning techniques to identify anomalies. Isolation Forest is used as the primary unsupervised anomaly detection algorithm, while Local Outlier Factor (LOF) and One-Class Support Vector Machine (One-Class SVM) are available for comparison. The framework supports real-time humidity monitoring using an Arduino Uno with a DHT11/DHT22 sensor and dataset-based analysis through CSV uploads. Data preprocessing, feature extraction, container-level aggregation, SQLite storage, and a Flask-based web dashboard are integrated into a single application. The system displays sensor readings and anomaly results and supports report generation. The proposed framework provides a lightweight approach to IoT sensor monitoring and anomaly analysis.