ML-Based Identification of Structured Light Schemes under Free Space Jamming Threats for Secure FSO-Based Applications

Abstract

This paper exploits for the first time the use of machine learning (ML) based techniques to identify complex structured light patterns under free space optics (FSO) jamming attacks for secure FSO-based applications. Five M-ary modulation schemes, construed using Laguerre and Hermite Gaussian (LG and HG) mode families, were used in this investigation. These include 8-ary LG, 8-ary superposition-LG, 16-ary HG, 16-ary LG and superposition-LG, and 32-ary LG and superposition-LG and HG formats. The work was conducted using experimental demonstrations for two different jammer positions. The convolutional neural network (CNN)-based ML method was utilized to differentiate between the stressed mode patterns. The experimental results show a 100% recognition accuracy for 8-ary LG, 8-ary superposition-LG, and 16-ary HG at 1, −2, and −2 dB signal-to-jammer ratios (SJR), respectively. For SJR values < 0 dB, the standard LG modes are the most affected by jamming and are not recommended for data transmission in such an environment. Besides, the accuracy of determining the jammer direction of arrival was investigated using CNN and a simpler classifier based on linear discriminant analysis (LDA). The results show that advanced networks (e.g., CNN) are required to achieve reliable performance of 100% direction determination accuracy, at −5 dB SJR, as opposed to 97%, at 2 dB SJR, for a simple LDA classifier.