Isolation forest enabled anomaly detection caused by internal and external disturbances in IsOWC system

Abstract An advanced approach for anomaly detection in inter-satellite optical wireless communication (IsOWC) system is proposed, utilizing the isolation forest (IF) algorithm to address challenges such as internal and external disturbances in the space environment. This algorithm aims to enhance detection reliability while minimizing false positives and negatives, thereby ensuring optimal system performance. A dataset containing 400 instances of relative intensity noise, propagation distance, and pointing error are used to train the model. Before training the model, a thorough exploratory data analysis is carried out to ensure effective data preparation. This involves visualizing feature importance values, creating box plots and kernel density estimates, and applying techniques like principal component analysis and t-distributed stochastic neighbor embedding. Additionally, the impact of the IF algorithm on different hyperparameters, bit error rate (BER), and optical signal-to-noise ratio (OSNR) is analyzed, demonstrating superior performance in anomaly detection with a receiver operating characteristic score of 97%, an accuracy of 93%, and a BER of -4.95 at OSNR of 24.76 dB. The proposed approach outperforms other techniques like one-class support vector machine and local outlier factor. These findings present a scalable, efficient, high-performance solution to anomaly detection in IsOWC system, which improves satellite network reliability, reduces operational costs, and enhances global connectivity. Furthermore, this approach has great potential to be adopted by leading space communication organizations as NASA, ISRO, SpaceX, OneWeb to revolutionize satellite-based communication infrastructures by monitoring anomalous behaviour between satellites and significantly improving the performance of deep space communication system.