Traditional inputs only authenticate identity at the initial lock screen. OrbitKeys leverages human micro-kinematics and hardware-isolated edge inference to continuously defend open app sessions against automated threats and unauthorized device takeovers.
OrbitKeys captures continuous touch parameters from the capacitive digitizer at 60Hz–120Hz. The pipeline maps flight lanes, swipe angular velocity, and inter-thumb span into an un-clonable kinetic fingerprint.
Static grids are vulnerable to coordinate-based screen recording. OrbitKeys uses Dynamic Spatial Shifting, making intercepted touch-logs mathematically useless.
No raw biometric data leaves the device. Inferences run locally via CoreML and TensorFlow Lite inside hardware-isolated partitions.
The edge engine computes an Anomaly Index Score (Aᵢ) in under 5ms.
Telemetry matches the verified baseline. Input stays fully unhindered.
Minor divergence detected. The framework silently steps up token security and injects cryptographic entropy to verify the session.
Threat threshold reached. Session tokens are revoked and Face ID / passcode is required before more input is accepted.
PASS = Safe Zone (Aᵢ < 0.3). WARN_CRYPT_ENTROPY = Warning Zone. BLOCK_STEP_UP_CHALLENGE = Breach isolation.