How greeshmita-23/sentinelrx works
SentinelRx is a notebook-based healthcare ML security project. Its documented workflow audits medication-adherence data, detects and removes target leakage, trains and evaluates models, then explores explainability, attacks, and defenses. The README documents these capabilities and a sequential workflow, but the notebook was not excerpted, so internal calls and implementation details cannot be verified; edges below reflect only documented relationships.
Main entry points: SentinelRx - Adversarially-Robust & Privacy-Preserving Medication-Adherence ML.ipynb Model evaluation · healthcareTrain.csv Training data · healthcareTest.csv Test data · secure_adherence_model.joblib Trained model
Components 21
Data integrity
- Training data
healthcareTrain.csv - Test data
healthcareTest.csv - Dataset audit
SentinelRx - Adversarially-Robust & Privacy-Preserving Medication-Adherence ML.ipynb - Leakage detection
SentinelRx - Adversarially-Robust & Privacy-Preserving Medication-Adherence ML.ipynb - Feature engineering
SentinelRx - Adversarially-Robust & Privacy-Preserving Medication-Adherence ML.ipynb
Model development
- Model training
SentinelRx - Adversarially-Robust & Privacy-Preserving Medication-Adherence ML.ipynb - Model benchmarking
SentinelRx - Adversarially-Robust & Privacy-Preserving Medication-Adherence ML.ipynb - Model calibration
SentinelRx - Adversarially-Robust & Privacy-Preserving Medication-Adherence ML.ipynb - Model evaluation
SentinelRx - Adversarially-Robust & Privacy-Preserving Medication-Adherence ML.ipynb - Feature importance
SentinelRx - Adversarially-Robust & Privacy-Preserving Medication-Adherence ML.ipynb - Trained model
secure_adherence_model.joblib
Security evaluation
- Adversarial examples
SentinelRx - Adversarially-Robust & Privacy-Preserving Medication-Adherence ML.ipynb - Membership inference
SentinelRx - Adversarially-Robust & Privacy-Preserving Medication-Adherence ML.ipynb - Label-flip poisoning
SentinelRx - Adversarially-Robust & Privacy-Preserving Medication-Adherence ML.ipynb - Backdoor simulation
SentinelRx - Adversarially-Robust & Privacy-Preserving Medication-Adherence ML.ipynb - Model extraction
SentinelRx - Adversarially-Robust & Privacy-Preserving Medication-Adherence ML.ipynb
Privacy and defense
- Randomized smoothing
SentinelRx - Adversarially-Robust & Privacy-Preserving Medication-Adherence ML.ipynb - Differential privacy
SentinelRx - Adversarially-Robust & Privacy-Preserving Medication-Adherence ML.ipynb - Poisoning sanitization
SentinelRx - Adversarially-Robust & Privacy-Preserving Medication-Adherence ML.ipynb - Runtime firewall
SentinelRx - Adversarially-Robust & Privacy-Preserving Medication-Adherence ML.ipynb
Other components
- Analyst
Connections 22
- Analyst → Dataset audit: runs notebook
README.md - Training data → Dataset audit: audits
README.md - Test data → Dataset audit: audits
- Dataset audit → Leakage detection: checks leakage
README.md - Leakage detection → Feature engineering: removes leakage
README.md - Feature engineering → Model training: prepares features
README.md - Model training → Model benchmarking: benchmarks classifiers
README.md - Model benchmarking → Model calibration: calibrates model
- Model calibration → Model evaluation: evaluates model
- Model training → Model evaluation: evaluates model
README.md - Model training → Trained model: saves model
- Model evaluation → Feature importance: explains predictions
README.md - Model evaluation → Adversarial examples: evaluates attack
README.md - Model evaluation → Membership inference: measures privacy attack
README.md - Model evaluation → Label-flip poisoning: evaluates attack
README.md - Model evaluation → Backdoor simulation: evaluates attack
README.md - Model evaluation → Model extraction: evaluates attack
README.md - Adversarial examples → Randomized smoothing: motivates defense
- Membership inference → Differential privacy: motivates defense
- Label-flip poisoning → Poisoning sanitization: motivates defense
README.md - Model training → Differential privacy: protects training
README.md - Model evaluation → Runtime firewall: guards runtime inputs
README.md