Generating a shared key between two parties from the wireless channel is an increasingly interesting topic. The process of obtaining information from the wireless channel is called channel probing. Previous key generation schemes probe the channel at a preset and constant rate without any consideration of channel variation or probing efficiency. To satisfy the usersâ requirements for key generation rate (KGR) and to use the wireless channel efficiently, we propose an adaptive channel probing scheme based on the proportional-integral-derivative controller, which is used to tune the probing rate. Moreover, we use the Lempel-Ziv complexity to estimate the entropy rate of channel statistics (received signal strength), which is considered as an indicator of probing efficiency. The experimental results show that the controller can dynamically tune the probing rate and, meanwhile, to achieve a user desired KGR. It stabilizes the KGR at the desired value with error below 1 bit/s. Besides, channel probing process is efficient under different user velocities, motion types, and sites.

Generating a shared key between two parties from the wireless channel is of increasing interest. The procedure for obtaining information from wireless channel is called channel probing. Previous works used a constant channel probing rate to generate a key, but they neither consider the tradeoff between the bit generation rate (BGR) and channel resource consumption, nor adjust the probing rate according to different scenarios. In order to satisfy users' requirement for BGR and to use the wireless channel efficiently, we first build a mathematical model of channel probing and derive the relationship between BGR and probing rate. Second, we introduce an adaptive channel probing system based on Lempel-Ziv complexity (LZ76) and Proportional-Integral-Derivative (PID) controller. Our scheme uses LZ76 to estimate the entropy rate of the channel statistics, e.g. the Received Signal Strength (RSS), and uses the PID controller to control the channel probing rate. Our experiments show that this system is able to dynamically adjust its probing rate to achieve a desired BGR under different moving speeds, different mobile types, and different sites. Our results also show that the standard deviation of the LZ76 calculator is less than 0.15 bits/s. The PID controller is able to stabilize the bit generation rate at a desired value with mean error of less than 0.9 bits/s.