Machine Learning-Based Vector Quantization for Secret Key Generation in Physical Layer Security

The growth of IoT in 6G networks necessitates efficient, lightweight security solutions for generating secure symmetric keys in low-power devices, as traditional public-key cryptography is inadequate due to complexity and overhead. In this paper, we propose a novel approach for physical layer key generation using channel reciprocity-based methods. Our contribution lies in the development of a unique feature-extraction method that reduces signal dimensions, enhances computational efficiency, and improves shared key accuracy. Additionally, we introduce an innovative Equalized Fuzzy C-Means (EFCM) vector quantization scheme that ensures clusters of equal size and maximizes entropy and randomness. Existing vector quantization schemes have communication overhead to address high Bit Error Rate (BER) through the exchanging of information over the channel. This can cause information leakage to a potential eavesdropper, which is addressed in our scheme by eliminating this communication overhead. Simulation results demonstrate the superiority of our approach, achieving higher key entropy, lower BER, and improved security compared to traditional scalar quantization and Principal Component Analysis-KMeans (PCA-KM) methods. The proposed scheme offers a promising solution for secure key generation in 6G networks and addresses the challenges posed by IoT devices' limited capabilities.