MACHINE LEARNING BASED PHYSICAL LAYER KEY GENERATION APPROACH FOR INTERNET OF THINGS.

Abstract

The communication device on the internet of things has limited resources such as bandwidth and memory. The limitation of resources cannot afford the computational cost of classical cryptography and suffers threats of information. Recently, several authors employed alternate ways of securing the internet of things using channel parameters of wireless communication. This paper proposes a channel parameter-based physical layer key generation approach. The proposed algorithm encapsulates other methods such as discrete wavelet transforms and machine learning algorithms. The employed discrete wavelet transforms methods reduce quantization errors and improve multi-bit formation. The employed clustering algorithm groups the bits in the form of blocks and 128 and 256 bits of keys. The proposed algorithm was simulated using MATLAB tools. For the evaluation of performance, estimate the ADR and length of the generated key. The estimation of ADR is based on the signal-to-noise ratio. The variation in noise changes the behaviours of the generated keys. The proposed algorithm compares with existing algorithms such as DCT, DWT, WPT, and BKQ. The performance of the results suggests that the proposed algorithm is very efficient compared to existing algorithms for physical layer key generation.