Finger Bending Biometrics (Smart Glove)
Sensor type - Resistive flex sensors (flexible potentiometers)
Sensor placement - One flex sensor mounted along the length of each finger
- Sensors aligned to capture finger bend and flex during typing
- Glove-based form factor ensures consistent placement across sessions
Sampling & data capture - Sampling rate: ~104 Hz
- Continuous time-series resistance values per finger
- Resistance increases proportionally with degree of finger bend
What was directly measured - Instantaneous finger bending dynamics (per finger, per time step)
What was inferred - User identity during PIN/password entry
- Finger coordination patterns, including non-typing fingers that move involuntarily during typing
Why this setup matters - Moves sensing below keystrokes and wrist motion
- Captures biomechanical intent, not just interaction outcomes
- Enables analysis of finger-level contribution and sensor count tradeoffs
Gamified Wearable Data Collection:Smartwatch IMU
Sensor type Sensor placement - Wrist-worn smartwatch
- Captures motion of the hand and wrist during touchscreen interactions
Sampling & data capture - Multi-axis motion data (linear acceleration + angular velocity)
- Continuous capture during:
- Gamified interactions
- Direct authentication pattern entry
What was directly measured - Wrist and hand movement trajectories during pattern execution
What was inferred - Authentication pattern similarity
- Feasibility of using gameplay-generated motion as training data
- Risk of inferring secret patterns via benign-looking apps
Why this setup matters - Demonstrates ecological validity: sensing during natural interaction, not lab-only tasks
- Highlights tradeoff between:
- Ease of deployment (wrist IMU)
- Loss of finger-level specificity (vs glove-based sensing)
Power Side-Channel Video Inference — Public Charging Hubs
Sensor type - External power measurement circuitry
- Measures voltage/current drawn during charging
Sensor placement - Embedded within or attached to public USB charging hubs
- No modification to the phone required
Sampling & data capture - Continuous power draw measurements during charging
- Captured while videos play on the smartphone
What was directly measured - Aggregate charging power consumption over time
What was inferred - Identity of videos being watched
- Effects of screen brightness, color dynamics, and device model
- User media consumption behavior
Why this setup matters - Turns infrastructure into a passive sensing adversary
- Exploits a signal users assume is non-informative
- Demonstrates that privacy leakage can occur without malware or permissions
