BHFVAL: Block chain-Enabled Hierarchical Federated Variational Auto encoder Framework for Secure Intrusion Detection...

Area of Science:

• Vehicular Networks
• Decentralized Learning
• Cybersecurity

Background:

• Centralized architectures in vehicular systems are vulnerable.
• Existing federated learning is inefficient in adversarial environments.
• Limited computational resources and varying network conditions pose challenges for secure data processing in IoV.

Purpose of the Study:

• To introduce an intelligent, effective, and secure learning model for the Internet of Vehicles (IoV).
• To address the vulnerability of centralized architectures and inefficiency of federated learning in adversarial environments.
• To enhance decentralized learning efficacy under limited computational resources and varying network conditions.

Main Methods:

• Blockchain-Enabled Hierarchical Federated Variational Autoencoder Learning (BHFVAL) model with a Reputation-Based Byzantine Fault Tolerance (RBFT) mechanism.
• Lightweight Edge-Computing (LEC) module for proximity-based processing to minimize communication latency.
• Osprey Optimization Algorithm (OOA) for hyperparameter optimization and Lightweight Secure Communication Protocol (LSCP) on Elliptic Curve-Based Homomorphic Encryption (ECHE) for secure V2X communication.

Main Results:

• UNSW-NB15 dataset achieved 96.83% accuracy and 96.65% F1-score (IID), and 95.74% accuracy and 95.40% F1-score (non-IID).
• CIC-IDS-2017 dataset achieved 97.36% accuracy, 97.2% AUROC, and 97.1% F1-score (IID), and 96.40% accuracy and 96.20% F1-score (non-IID).
• Demonstrated strong detection performance and adaptability in decentralized vehicular networks.

Conclusions:

• The BHFVAL framework is dependable, adaptable, and effective for decentralized privacy-sensitive vehicular networks.
• The model successfully enhances security and learning efficacy in challenging IoV environments.
• The proposed methods significantly improve data processing security and learning efficiency in vehicular systems.