An improved approach: adaptive learning for high-speed data encryption of low-earth orbit (LEO) satellites

This paper examines the critical problem of ensuring data security in low-Earth orbit (LEO) satellite communication systems, where resources are limited, and the risk of cyberattacks is heightened. An innovative approach to continuous user authentication is proposed, based on adaptive machine learning using logistic regression and support vector machines, combined with robust cryptographic protocols. The study's primary goal is to develop a lightweight and efficient encryption method that guarantees a high level of data security without significantly increasing the computational load on LEO satellite onboard systems. The article analyzes existing continuous authentication methods and identifies their vulnerabilities, such as dependence on specific features, scalability issues, and susceptibility to targeted attack scenarios. The proposed approach is based on the adaptive sliding window method, which allows dynamic adaptation to changes in user behavior and effectively detects anomalies indicating potential unauthorized access attempts. To safeguard data, the study proposes using functional encryption and decentralized key generation, enhancing the system's resilience to various types of attacks. Preliminary simulation results using mouse movement data demonstrate that the proposed approach achieves high anomaly detection accuracy (over 80%) with minimal computation time (less than 9 ms). This method can be applied to protect data in various communication systems with LEO satellites, including flight control, telemetry, and data transmission systems.