A Federated Learning Secure Encryption Framework for Autonomous Systems

Venkatraman Balasubramanian, Moayad Aloqaily, Mohsen Guizani

Abstract: As autonomous vehicles, smart infrastructure, and connected devices become integral components of our daily lives, the need to protect communication and establish trust among entities is fundamental. This problem necessitates the development of robust authentication mechanisms tailored for autonomous systems. In this paper, we address the critical challenge of designing authentication methods to verify the legitimacy of messages and participants in autonomous systems. The proposed methods aim to validate the origin of messages and the identity of participants, ensuring that only authorized entities interact with the system. Achieving this requires cryptographic techniques, digital signatures, and secure key management. To that end, we developed an algorithmic framework that includes message digest generation, digital signature creation, and recipient-side verification. Additionally, participant authentication and message integrity checks are incorporated to fortify the authentication process. The algorithm leverages public key cryptography to verify digital signatures and ensure the message's integrity. Second, we develop a simulation by harnessing Federated Learning (FL) which provides a dynamic and self-improving authentication mechanism that aligns with the high-reliability demands of modern autonomous applications on MATLAB. We elaborate on how addressing this problem is essential to bolster the security of autonomous systems, safeguard against cyber threats, and instill trust in the reliability and authenticity of communication within these systems. The proposed framework has been tested and validated on MATLAB.