Puzzle: A Shape-based Secret Sharing Approach By Exploiting Channel Reciprocity in Frequency Domain

In this paper we propose a shape-based approach in frequency domain to extract a shared key by exploiting the observation that wireless channel is reciprocal due to multi-path fading. Unlike the traditional quantization approach in time domain, no training sequences or predetermined pulses are needed to be transmitted in our approach. The correlated power spectral density of the transmitted packets served as the common random source between Alice and Bob. We use Lowess smoothing to mitigate measurement errors and interference and then use pattern matching to encode the shape of the spectrum. We name the technique as Puzzle for secrets are generated by finding right pieces(shape patterns) and then putting them together. Implementation in softwaredefined radios (SDR) demonstrates the feasibility of extracting a 6-bit secret per measurement with an average bit mismatching rate 5% in a 20 MHz band. Experiments show that with eavesdropper near by, the leaked information of each secret bit generated by Puzzle is about 0.05 bit, which is low in comparison with a RSSI-based method ASBG.