We take security seriously for NeuralFlight. The following versions are currently supported with security updates:
| Version | Supported |
|---|---|
| 2.0.x | ✅ |
| < 2.0 | ❌ |
Please do not report security vulnerabilities through public GitHub issues.
If you discover a security vulnerability in NeuralFlight, please report it privately to help us address it before public disclosure.
- Email: Send details to
[email protected] - Subject: Include "SECURITY" in the subject line
- Details: Provide the following information:
- Type of vulnerability
- Affected component(s)
- Steps to reproduce
- Potential impact
- Suggested fix (if any)
- Acknowledgment: Within 48 hours of your report
- Updates: Regular status updates every 5 business days
- Timeline: We aim to address critical vulnerabilities within 7 days
- Credit: We will acknowledge your contribution in the fix announcement (unless you prefer to remain anonymous)
NeuralFlight is designed for research and educational purposes. When using this software:
- ✅ DO: Use in controlled environments (simulation, testing areas)
- ✅ DO: Test thoroughly before any real-world applications
- ✅ DO: Follow local regulations for drone operation
- ❌ DON'T: Use in safety-critical applications without extensive validation
- ❌ DON'T: Rely solely on BCI control for autonomous systems
- ❌ DON'T: Operate real drones near people or property without proper safeguards
- EEG Classification: 73% accuracy means ~27% error rate - not suitable for critical applications
- Real-time Performance: Latency exists between thought and action
- Subject Variability: Performance varies significantly between individuals
- Environmental Factors: EEG signals are affected by electrical interference, muscle artifacts, and environmental noise
If adapting for real drones:
- Implement multiple redundant safety systems
- Always include manual override capability
- Use geofencing and altitude limits
- Test extensively in simulation before hardware deployment
- Follow manufacturer safety guidelines
We regularly monitor our dependencies for known vulnerabilities:
- PyTorch: We track CVEs and update when security patches are released
- OpenCV: Updated regularly for security fixes
- Mediapipe: Google maintains this library with regular security updates
- MNE-Python: Community-maintained with security considerations
You can check for vulnerable dependencies using:
pip install safety
safety check -r requirements.txtNeuralFlight processes EEG data locally:
- ✅ No data is transmitted to external servers
- ✅ All processing happens on your local machine
- ✅ You control data storage and deletion
- ✅ PhysioNet dataset usage follows their terms of service
- We do not collect any personal information
- Webcam data (for gesture control) is processed locally only
- No telemetry or analytics are collected
- No third-party tracking
If you're contributing code:
- Input Validation: Always validate user inputs
- Error Handling: Handle exceptions gracefully
- Dependencies: Only add well-maintained, security-vetted packages
- Sensitive Data: Never commit API keys, credentials, or personal data
- Code Review: All PRs undergo security review before merging
We will announce security updates through:
- GitHub Security Advisories
- Release notes
- README notifications
Subscribe to repository notifications to stay informed.
This security policy covers:
- ✅ The core NeuralFlight codebase
- ✅ Official demo applications
- ✅ Documentation and examples
- ❌ Third-party integrations (report to those projects)
- ❌ User-modified versions (your responsibility)
For general security questions (not vulnerability reports), open a GitHub Discussion or contact us at [email protected].
Remember: NeuralFlight is an experimental research platform. Always prioritize safety when working with autonomous systems!