Inspiration
Healthcare is one of the fields where artificial intelligence can make a meaningful impact, but I believe that intelligent healthcare systems must be both efficient and secure.
I was inspired to develop SurgeGuard-Agent after exploring the challenges of surgical risk monitoring, hospital inventory management, and patient data protection.
My original idea was to build a healthcare workflow dashboard that could identify unusual simulated sensor conditions and monitor medical supplies. As the project evolved, I realized that healthcare automation also raises an important question: How can an AI-driven workflow respect patient consent and prevent unauthorized access to sensitive information?
This inspired me to extend SurgeGuard-Agent with a consent-aware security verification module, combining operational monitoring with responsible AI security principles.
What it does
SurgeGuard-Agent is an educational healthcare workflow prototype that demonstrates three core capabilities:
1. Surgical Risk Monitoring
Analyzes simulated movement measurements and sensor conditions to generate rule-based alerts when predefined risk conditions are detected.
2. Medical Inventory Monitoring
Checks simulated medical supply availability and identifies potential stock shortages based on requested quantities.
3. Consent-Aware Security Verification
Evaluates simulated healthcare data access requests using three security checks:
- Patient identifier consistency.
- Authorized consent scope.
- Consent expiration.
The system generates ALLOW or BLOCK decisions with explanations, helping demonstrate how unauthorized requests can be rejected before access is approved.
An interactive Streamlit dashboard displays monitoring results, security decisions, and downloadable CSV audit records.
How we built it
I developed SurgeGuard-Agent independently using Python, Streamlit, Visual Studio Code, GitHub, and Dify.
- Python: Implemented risk monitoring, inventory verification, and security rules.
- Streamlit: Developed the interactive healthcare monitoring dashboard.
- Dify: Explored AI workflow orchestration and API integration during earlier development.
- Pandas: Organized simulation results and generated CSV evaluation reports.
- GitHub: Managed the source code, project documentation, and public repository.
The project follows a modular architecture with separate Python files for monitoring and security verification.
The security_guard.py module checks patient identifiers, consent scopes, and expiration timestamps using deterministic rules. The test_security.py script evaluates security scenarios, while combine_reports.py consolidates the evaluation results.
The deployed dashboard can operate using local Python rules without a configured Dify API key.
Challenges we ran into
1. Integrating different technologies
Connecting Python modules, Streamlit components, Dify workflow experiments, and evaluation reports required debugging and repeated testing.
2. Designing reliable security checks
One major challenge was ensuring that untrusted instruction text could not directly override authorization decisions. I addressed this by separating consent verification from instruction interpretation.
3. Expanding the original project
Transforming a healthcare monitoring dashboard into a consent-aware security prototype required additional modules, testing, and interface changes.
4. Understanding real-world limitations
I learned that successful demonstration tests do not automatically prove that a system is secure against sophisticated attacks. The prototype still requires stronger authentication, consent verification, and broader adversarial testing.
Accomplishments that we're proud of
I am particularly proud of developing and publishing SurgeGuard-Agent as an independently built educational prototype.
My key accomplishments include:
- Developed an interactive healthcare monitoring dashboard.
- Implemented a separate consent-aware security verification module.
- Successfully completed 5 out of 5 configured security tests.
- Successfully completed 4 out of 4 healthcare simulation scenarios.
- Generated a combined evaluation dataset containing 9 records.
- Added transparent ALLOW/BLOCK decisions with explanations.
- Implemented downloadable CSV evaluation reports and security audit logs.
- Published the project on GitHub and deployed the dashboard using Streamlit.
These accomplishments demonstrate the expected behavior of the implemented rules in selected scenarios, rather than production-level clinical or cybersecurity performance.
What we learned
Developing SurgeGuard-Agent helped me gain practical experience in software development, healthcare workflow simulation, and AI security.
I learned how to:
- Build modular Python applications.
- Design interactive interfaces using Streamlit.
- Experiment with workflow automation using Dify.
- Implement deterministic consent-based access checks.
- Test healthcare information access scenarios.
- Generate reproducible evaluation reports.
- Understand the importance of trust boundaries and security assumptions.
- Recognize the difference between a functional prototype and a production-ready security system.
My biggest learning was that responsible healthcare AI requires more than intelligent automation. It also requires transparency, authorization, accountability, and respect for patient privacy.
What's next for SurgeGuard-Agent
I plan to improve SurgeGuard-Agent by:
- Connecting simulated AI agent tool requests directly to the consent-verification system.
- Introducing stronger identity authentication and consent validation mechanisms.
- Exploring real-time consent revocation and more detailed audit logging.
- Expanding adversarial security testing and performance evaluation.
- Improving the dashboard with additional monitoring scenarios and clearer explanations.
- Exploring standards-based healthcare interoperability in safe, fictional testing environments.
My long-term vision is to develop SurgeGuard-Agent into a more comprehensive research prototype for secure and accountable healthcare automation.
GitHub Repository: https://github.com/preethidavidson249140/SurgeGuard-Agent
Live Demo: https://surgeguard-agent-bafgymxyjjjd5vmctrbmo3.streamlit.app/
Disclaimer: SurgeGuard-Agent is an educational prototype using fictional data. It is not a clinical decision-support system and does not connect to real patient records or live ABDM/ABHA infrastructure.
Log in or sign up for Devpost to join the conversation.