Cybersecurity of Sensor Systems for State Sequence Estimation A Machine Learning Approch

Modern sensor systems deployed in smart grids and critical infrastructure are increasingly vulnerable to cyber threats that can disrupt state sequence estimation and compromise system reliability. This project presents a machine learning–driven framework to enhance cybersecurity by detecting malicious network activities in real time. The system integrates data preprocessing, feature selection, and supervised learning techniques to classify network traffic as benign or attack. Multiple algorithms, including Random Forest, XGBoost, Extra Trees, LightGBM, and K-Nearest Neighbors, are implemented and evaluated to ensure robust detection performance across diverse scenarios.

To further improve accuracy and resilience, a hybrid model is developed by combining the strengths of multiple algorithms, resulting in enhanced predictive capability and generalization. A web-based application built using Flask provides an interactive interface for dataset upload, model selection, and real-time prediction. The framework also generates actionable security recommendations based on prediction outcomes, helping mitigate threats such as denial-of-service attacks and unauthorized intrusions. Overall, the proposed approach delivers a scalable, efficient, and adaptive solution for securing sensor-driven environments in real-world applications.

Keywords: Cybersecurity, Sensor Systems, Machine Learning, Intrusion Detection, State Sequence Estimation, Smart Grid Security, Classification Models, Anomaly Detection.