Behavioural sensor data as randomness source for IoT devices

Random numbers generators are very important for cryptography and using biometrics as a source for randomness is on ascending trend. Because of the multitude of IoT devices present in our daily lives the need for keys and sources of randomness increased. This paper assesses the feasibility of using IoT devices to gather data on human behaviour for creating randomness. This paper analyses the suitability of using bio- metric data collected from 6 smartphone sensors: accelerometer, gravity, gyroscope, linear acceleration, magnetometer, rotation, and sound. The sensor data is run through all three well-known batteries of tests: NIST, ENT and Dieharder. We demonstrate that human gait is highly predictable and shouldn’t be used as a source of randomness. We also show that contrary to popular believe using data from two sensors slightly improves randomness, but it still doesn’t pass all tests.