High entropy secrecy generation from wireless CIR

The physical characteristics of wireless channels, in particular the channel impulse response (CIR), have become an attractive source of shared secrecy between two communicating parties for many wireless applications. However, ensuring high-entropy secret bits from the CIR may not be always feasible, including in IoT and/or indoor environments, where the coherence time may not be easy to obtain, or mobility is limited. Many techniques have been developed to resolve the above problems, including privacy amplification for higher entropy output, and pre-processing for data distillation. In this paper, we develop a framework for extracting entropy from sequentially arriving channel impulse response observation samples, and propose algorithms for both online and offline versions of the problem. We fully characterize the optimal solutions to these problems through a series of analytic results, and establish the optimality of our proposed online and offline algorithms, namely A_EMSON and A_EMSOFF to solve the problem. We also verify the performance of the algorithms through comparisons with the existing related techniques, and evaluate their capability to ensure high-entropy under different channel models using MATLAB simulations.