Key Gestures:
Gesture-Based Password Authentication
Team: Camryn Garrett and Ryan Milligan
I. Project
Traditional keyboard-based password input is not accessible for everyone
Key gestures solves this by:
- Utilizing easy password creation/entry for those lacking finger dexterity
- Requiring no extraneous hardware
- Providing a user friendly and low maintenance option
II. System
Functional Block Diagram
Software Design Diagram
III. Methods
Reading User Input
- 58 elements – each element belonging to a single key
- Each element given score from 0 to 1 representing sequence
Neural Network Processing
- Input: 58 keys
- Output: 10 gestures
- Pytorch implementation
- Trained with 9000 inputs
Graphical Display
- Demonstration and password mode
- Real-time display
- User customization
IV.Conclusion
Results
- Accepts and correctly interprets inputs from all QWERTY keyboard layouts
- Recognizes 10 gestures with 93.7% accuracy
Future Work
- Full OS integration
- More accurate gesture representation
- Custom gesture encryption
