Machine Learning-Assisted Secure Random Communication System by Deploying Knn-Based Receiver

Enhancing physical layer security of communication systems provides an additional strength against potential eavesdropping. Many conventional communication architectures have already been incorporated with machine learning algorithms which has uplifted the performance and security to an unprecedented level. With this motivation, we propose an unconventional machine learning assisted random communication system (ML-RCS). It utilizes the skewed alphastable noise as a random carrier to covertly convey the binary information. The proposed ML-RCS employs a transmitter and a designated kNN-based receiver. The pretraining of the designated receiver has been done by privately compiled dataset and fixed key-the pulse length- to securely extract the concealed binary bits. The Monte Carlo simulations (single-bit transmissions and confusion matrices) reflect enhanced performance in terms of data rate. Conversely, the false negative rate of 93.8% reflects the eavesdropper's inability to intercept the communications. Therefore, with built-in physical layer security, ML-RCS can emerge as a promising candidate for uplifting the defense of next-generation communication devices.