Drinking water quality has long been a source of worry. In order to ensure that the water delivered by utilities is safe for human use, independent laboratories have traditionally examined the water. SCADA systems play a central role in monitoring and controlling water treatment facilities, overseeing and regulating a multitude of processes. The analysis of sensor data from multiple sites involves setting alarms to detect abnormalities caused by potential spoofing in communication between SCADA servers and PLCs, which could result in alterations to sensor and actuator readings. This study presents a strong cybersecurity solution by utilizing Long Short-Term Memory(LSTM), and performing clustering to classify the unlabeled attacks based on their characteristics. Developing a simulation model of the water treatment process aids in achieving a thorough understanding and evaluating the system’s functionality and performance across different operational scenarios. The system generates reports enabling users to take recommended actions.

The objectives of this project are:

• Enable detection of cyber-physical attacks aiming to manipulate sensor readings, introduce harmful chemicals, or alter pH levels in the water supply.

• Enhance the security infrastructure to safeguard public health by promptly identifying and minimizing potential threats arising from cyber-physical attacks on water treatment systems.

• Develop and implement an unsupervised clustering framework specifically designed to classify unlabeled attacks based on their characteristics, enabling more informed and targeted responses for each identified cluster.

• To simulate the six-stage process, Enabling a thorough evaluation of the functionality and performance of the water treatment system across various operational situations.

• To visually highlight detected attacks, we’ll use data analysis tools to create easy-to-understand graphs and reports of sensor readings in the water treatment system.

• Using a Real-life dataset (SWaT: Secure Water Treatment) created for research and testing purposes in the field of cybersecurity for water treatment systems.

• Ensure public health safety by detecting anomalies in the water treatment process.

• Applying preprocessing on operational data by cleaning, sampling, and scaling the data for better accuracy, and employing feature selection to curate a subset of sensors that are precisely targeted for analysis.

• The project involves utilizing a deep learning model based on LSTM (Long Short-Term Memory) to detect attacks within the water treatment system, utilizing a labeled dataset that categorizes instances as either ’attack’ or ’normal’ based on sensor readings.

• To optimize the approach, the project will utilize clustering techniques to classify unlabeled attacks based on distinct characteristics, enabling customized actions for each identified attack within the water treatment system.

• To facilitate comprehension and emphasize the detected attacks, the project will use data analysis tools to visualize sensor readings. This visualization will include reports and graphs generated by the data analysis tools, aiding in the easy interpretation of the data and highlighting identified attacks within the water treatment system.

• To simulate the six-stage process, the project will utilize simulation techniques, allowing for a comprehensive understanding and assessment of the entire water treatment system’s functionalities and performance within diverse operational scenarios.