Perfect Attackability Detection for Linear Systems with Deep Learning

This project seeks to find a solution for determining perfect attack matrices for LTI output feedback systems. It utilizes various NN models to accomplish this task.

File Structure:

• src: Source code
  • attackability_model.py: Core function for training the model, called from run.py
  • model.py: NN model definition
  • optimal.py: Classic numerical optimization to compare NN model to
  • run.py: Generates inputs for the model and runs the training
  • ss_loss.py: Loss function
  • state_space_generator.py: Data generator for state space models to train/test/evaluate on
• config.yaml: Model training parameters
• model.pth: Saved model