Utilization of adaptive techniques for channel estimation and intrusion detection in OWC systems.

Autor: Abd-Elgawad, Lamiaa A., Sallam, Youssef F., Hussein, Gamal A., Elabyad, Gaber S. M., Oraby, Osama A., El-Hag, Noha A., Ahmed, Hossam Eldin H., Saleeb, Adel A., El-Bahnasawy, Nirmeen A., Hammad, Randa S., El-Bendary, Mohsen A. M., El-Samie, Fathi E. Abd
Zdroj: Journal of Optics (09728821); Jul2024, Vol. 53 Issue 3, p1709-1721, 13p
Abstrakt: Sustaining channel equalization effectiveness, when using Optical Wireless Communication (OWC) systems is a challenging issue. In order to deal with the physical deficiencies in the Free Space (FS) channel, different adaptive channel estimation techniques are investigated in this study. Standard Least Mean Square (LMS), and LMS with Activity Detection Guidance (ADG) and Tap Decoupling (TD) are the adaptive channel estimation techniques taken into consideration. During the channel estimation process, random inputs are used, including both white and colored inputs. According to the simulation results, the channel estimation issue with white inputs can be resolved using the LMS algorithm with ADG. On the other hand, the TD is necessary for accurate channel tap estimation for colored inputs. Since the step size, number of estimated taps, and noise variance affect the asymptotic error performance of the channel estimators, the suggested technique is evaluated with unit noise variance, 0.07 noise variance, and 0.01 noise variance with a fixed correlation factor. When the correlation factor is equal to 0.1, 0.5, or 0.9 with the noise variance fixed, the effects of varying the correlation factor and the number of iterations on the estimator ability to remove channel distortion are explored. As a result, we become able to enhance the suggested channel estimation and obtain its best asymptotic performance and convergence rate. Transferring data without taking security into account is quite risky. Hence, we introduce an anomaly-based Intrusion Detection System (IDS) model that depends on Convolutional Neural Network (CNN) using Visual Geometry Group with 16 layers (VGG16). Our model achieved high levels of accuracy, and attack detection rates of 99.92%. [ABSTRACT FROM AUTHOR]
Databáze: Complementary Index
Externí odkaz: