E-DIM: Federated Learning-Based Secure Autonomous Communication Through Enhanced Data Integrity Monitoring
Balasubramanian, Venkatraman Aloqaily, Moayad Guizani, Mohsen Ouni, Bassem
Balasubramanian, Venkatraman
Aloqaily, Moayad
Guizani, Mohsen
Ouni, Bassem
Supervisor
Department
Machine Learning
Embargo End Date
Type
Journal article
Date
License
Language
English
Collections
Research Projects
Organizational Units
Journal Issue
Abstract
As self-driving vehicles, smart infrastructure, and interconnected devices seamlessly become part of our daily existence, a pivotal challenge emerges—establishing confidence and ensuring secure communication. Thus, in this article, we address two main problems: first, the vulnerabilities associated with operational cybersecurity in autonomous aerial vehicle (AAV)-assisted vehicular and wireless sensor networks. The increasing use of AAVs in these networks leads to heightened operational cybersecurity vulnerabilities. Addressing these vulnerabilities is critical to prevent potential catastrophic harm and malicious attacks on these networks; and second, robust authentication in autonomous systems. This involves ensuring secure communication and establishing trust among entities within the rapidly evolving landscape of self-driving vehicles, smart infrastructure, and interconnected devices. This challenge necessitates not only selecting the most trusted nodes but also ensuring secure communication. Both issues necessitate innovative solutions to bolster security and resilience in these technological domains. To that end, we propose enhanced data integrity monitoring (E-DIM), an enhanced data integrity monitoring framework that relies on the packets transmitted to the central server. Each device generates an authentication parameter that is communicated to the central server for verification of the packet’s integrity. E-DIM does not use expensive encryption algorithms, thereby reducing the overall size of the authentication message code (AMC). We show how we modify E-DIM to be used as a hierarchical federated learning framework. E-DIM maximizes the probability of selecting the best node among candidate nodes to support autonomous vehicles. Finally, we provide proof-of-concept performance analysis using MATLAB to illustrate how E-DIM performs better than random node selection, which is one of the popular state-of-the-art models.
Citation
V. Balasubramanian, M. Aloqaily, M. Guizani, B. Ouni, "E-DIM: Federated Learning-Based Secure Autonomous Communication Through Enhanced Data Integrity Monitoring," IEEE Transactions on Computational Social Systems, vol. PP, no. 99, pp. 1-12, 2026, https://doi.org/10.1109/tcss.2025.3648193.
Source
IEEE Transactions on Computational Social Systems
Conference
Keywords
46 Information and Computing Sciences, 4604 Cybersecurity and Privacy, 4605 Data Management and Data Science, 4606 Distributed Computing and Systems Software
Subjects
Source
Publisher
IEEE