- Extensive analysis and understanding of the New Gas Pipeline Dataset
- Attacking rule based models using explainable AI techniques
- ZOO based coordinated descent algorithm for handling binary and continuous features (+ one hot encoder + constraints + attack property preservation)
- Analyse defense mechanisms (Model based and data based)
- data: Contains the raw dataset and the processed data used for training the deep learning models
- src: Contains the source code for pre-processing the data and performing exploratory data analysis
- results: Contains the results of the hyper parameter search and model comparison
- ICS_IDS: Performs various adversarial attacks (Iterative FGSM and ZOO) and also evalutes the effectiveness of different defense strategies.
For more information about the project: final_report