An Evaluation On The Entropy Supplying Capability Of Smartphone Sensors.

This article focuses on evaluating the entropy supply capability of unprivileged smartphone sensors as noise sources for random number generation, using min-entropy estimation methods from NIST SP 800-90B. The study comprehensively analyzes sensor data from 300 Android and 30 iOS smartphones across various models and operating system versions, simulating both stationary and mobile working states. Results indicate that Android ACCELEROMETER sensors provide at least 1.5 bits of entropy per data acquisition in stationary states, while iOS GYROSCOPE sensors yield at least 20 bits, with entropy generally increasing in mobile states. Correlation analyses suggest that random numbers generated from these sensors are difficult to predict even by devices in similar environments. The paper concludes with design recommendations for sensor-based random number generators, emphasizing the use of ACCELEROMETER sensors on Android and GYROSCOPE sensors on iOS due to their stable entropy output.