How peerahammad/pard-in-network-traffic-using-switching-state-space-models works
PARD-SSM is a research system for inferring changing network-attack regimes and hidden states from traffic sequences. The documented workflow loads network data, engineers normalized and windowed observations, applies state-space models and inference methods, then evaluates and visualizes regime estimates. The graph preserves the documented explainability and experiment capabilities. Several links reflect README-level architecture because their implementation details were not sampled; those are cited to the README rather than inferred from module co-location.
Main entry points: run_baseline.py Baseline experiments · feature_engineering.py Feature engineering · run_switching.py Switching experiments · switching_ssm.py Switching SSM
Components 19
Traffic preparation
- Dataset loader
src/data_processing/dataset_loader.py - Feature engineering
src/data_processing/feature_engineering.py
State-space models
- Linear SSM
src/models/linear_ssm.py - Nonlinear SSM
src/models/nonlinear_ssm.py - Switching SSM
src/models/switching_ssm.py
Inference methods
- Kalman filter
src/inference/kalman_filter.py - Extended Kalman filter
src/inference/ekf.py - Unscented Kalman filter
src/inference/ukf.py - Variational switching
src/inference/variational_switching.py
Evaluation and analysis
- Regime metrics
src/utils/metrics.py - Result visualization
src/utils/visualization.py - SHAP explainability
src/explainablity/shap_explainer.py
Research experiments
- Baseline experiments
experiments/run_baseline.py - Switching experiments
experiments/run_switching.py - Baseline comparisons
experiments/baselines.py - Ablation studies
experiments/ablation_study.py - Evaluation metrics
experiments/evaluation_metrics.py - Significance analysis
experiments/statistical_significance.py
Other components
- Network traffic
Connections 23
- Network traffic → Dataset loader: loads data
README.md - Dataset loader → Feature engineering: provides datasets
README.md - Feature engineering → Linear SSM: builds observations
README.md - Feature engineering → Switching SSM: builds observations
README.md - Linear SSM → Kalman filter: builds filter
src/models/linear_ssm.py - Nonlinear SSM → Extended Kalman filter: builds filter
src/models/nonlinear_ssm.py - Nonlinear SSM → Unscented Kalman filter: builds filter
src/models/nonlinear_ssm.py - Switching SSM → Variational switching: infers regimes
README.md - Baseline experiments → Feature engineering: preprocesses data
experiments/run_baseline.py - Baseline experiments → Linear SSM: runs baselines
experiments/run_baseline.py - Baseline experiments → Nonlinear SSM: runs baselines
experiments/run_baseline.py - Baseline experiments → Regime metrics: computes metrics
experiments/run_baseline.py - Baseline experiments → Result visualization: plots results
experiments/run_baseline.py - Switching experiments → Feature engineering: preprocesses data
experiments/run_switching.py - Switching experiments → Variational switching: runs inference
experiments/run_switching.py - Switching experiments → Regime metrics: computes metrics
experiments/run_switching.py - Switching experiments → Result visualization: plots results
experiments/run_switching.py - Baseline comparisons → Baseline experiments: compares methods
- Ablation studies → Switching SSM: tests variants
- Evaluation metrics → Regime metrics: uses metrics
experiments/evaluation_metrics.py - Evaluation metrics → Result visualization: plots evaluations
experiments/evaluation_metrics.py - Significance analysis → Evaluation metrics: tests results
- Switching SSM → SHAP explainability: supports explanations
README.md