Date of Graduation
Summer 2026
Degree
Master of Science in Computer Science
Department
Computer Science
Committee Chair
Razib Iqbal
Abstract
The increasing number of Internet of Things (IoT) devices in smart environments, like smart homes, smart offices, and smart classrooms, creates a growing demand for seamless automation and synchronization among them. Accurately identifying and grouping related sensors is a key step for creating the automated policies that control actuators. However, this process is often hindered by anomalous data in sensor readings, which can obscure the sensor relationships identified during the grouping process. These anomalies can be due to real-time intrusions or irregularities identified in past data that occurred over a period. These anomalies compromise the effectiveness and reliability of the automation policies generated for managing smart environments. In this research project, I investigated both types of anomalies affecting sensor automation in the smart environment. Firstly, I propose a new method for detecting historical anomalies that calculates the total number of sensor events within a time window and then uses unsupervised learning for detection. Then, for real-time anomalies, I propose a novel unsupervised graph neural network model that uses a fixed sensor-relationship graph to forecast next-step sensor readings, flagging deviations from expected behavior as anomalies in real time. Results from these methodologies across three custom datasets and one public dataset demonstrate promising performance in detecting and removing relevant anomalies. I evaluated the effectiveness of both methods by leveraging existing sensor inference techniques and comparing how these anomaly detection methods improve smart environment automation.
Keywords
multivariate time series data, sensor frequency, adjacency matrix, clustering scores, isolation forest, graph convolution network, gated recurrent unit
Subject Categories
Other Computer Engineering
Copyright
© Md Asif Tanvir
Recommended Citation
Tanvir, Md Asif, "Anomaly Detection in Smart Iot Environments Using Unsupervised Event Filtering and Graph-Based Forecasting" (2026). Graduate Theses/Dissertations. 4200.
https://bearworks.missouristate.edu/theses/4200
