Federated Learning-Based Predictive Traffic Management Using a Contained Privacy-Preserving Scheme for Autonomous...

Area of Science:

• Intelligent Transport Systems (ITS)
• Autonomous Vehicle (AV) communication networks
• Machine Learning for traffic management

Background:

• Traditional traffic management systems risk user privacy and data security.
• Real-time data handling in vehicles exposes sensitive information to adversaries.
• Existing models lack robust privacy preservation and security against sophisticated attacks.

Purpose of the Study:

• To introduce a Federated Learning-based Predictive Traffic Management (FLPTM) system for Autonomous Vehicles (AVs).
• To enhance service access and privacy within Intelligent Transport Systems (ITS).
• To mitigate adversarial threats and ensure data integrity in vehicle communication networks.

Main Methods:

• Implementation of a Contained Privacy-Preserving Scheme (CPPS) for decentralized data processing.
• Utilizing Federated Learning (FL) for collaborative model training without raw data sharing.
• Integrating classifier-based learning, state modeling, and access permissions for enhanced security.

Main Results:

• Reduced communication costs by 23.29% through Federated Learning.
• Mitigated adversarial effects by 16.1% using the CPPS framework.
• Improved access time efficiency by 18.95% in the proposed system.

Conclusions:

• The FLPTM system effectively optimizes traffic management and enhances privacy for AVs.
• The CPPS framework provides robust security against man-in-the-middle attacks and data breaches.
• Federated Learning significantly improves security, reduces costs, and enhances efficiency in ITS environments.