Machine learning–based detection of cyber attacks in smart grid and SCADA network traffic.
Smart grid and SCADA networks rely on heterogeneous and resource-constrained communication protocols such as MODBUS and DNP3. These protocols are often vulnerable to cyber-attacks due to limited security mechanisms, which can impact grid reliability and demand response programs.
Traditional rule-based security systems are insufficient to detect complex and evolving attack patterns in cyber-physical systems.
To design, implement, and evaluate machine learning and deep learning models for detecting intrusions and anomalous behavior in smart grid and SCADA network traffic.
- Data preprocessing and feature engineering on real-world and benchmark datasets
- Supervised learning models: Random Forest, Decision Trees, Support Vector Machines
- Unsupervised learning models: k-means clustering
- Deep learning models: Feed-Forward Neural Networks, LSTM
- Comparative evaluation of models across multiple datasets and attack scenarios
- KDD Cup 99
- NSL-KDD
- MODBUS network traffic
- DNP3 network traffic
- Gas Pipeline dataset (Mississippi State University)
- Machine learning and deep learning models achieved improved detection performance compared to traditional baseline approaches
- LSTM models demonstrated strong capability in detecting sequential and time-dependent attack patterns
- Ensemble-based models provided robust performance across heterogeneous datasets
- Programming: Python
- ML/DL: Scikit-learn, TensorFlow / PyTorch
- Data Processing: Pandas, NumPy
- Environment: Linux, HPC environments
data/ -> datasets
src/ -> source code
results/ -> experiment outputs
assets/ -> figures and plots
- Experiments were conducted using publicly available and benchmark datasets
- Data preprocessing, feature extraction, and model training steps are documented
- Random seeds and model configurations are configurable for repeatability
- Results reported are based on multiple experimental runs
pip install -r requirements.txt
python src/main.py
This work was conducted as part of my PhD research in Computer Engineering, focusing on applied machine learning for energy systems and cyber-physical system security.
Related work from this project has been published in peer-reviewed conferences and journals in the areas of smart grids, cybersecurity, and applied machine learning. Related peer-reviewed publications are listed in my CV.
- Applied machine learning
- Cyber-physical system security
- Data preprocessing and feature engineering
- Experimental evaluation
- Reproducible research workflows
Rajesh Manicavasagam
Machine Learning Engineer | PhD in Applied Machine Learning
GitHub: https://github.com/rmanicav