EPDA: An Efficient and Privacy-preserving Dual Aggregation Scheme for Federated Learning in VANETs
Abstract
Abstract Federated Learning (FL) enables devices to collaboratively train models without sharing raw data, but attackers can still infer sensitive information from uploaded updates. Existing cryptographic methods provide security but often lack essential privacy properties such as anonymity and unlinkability, while introducing high computational and communication overheads. These challenges limit the deployment of efficient, privacy-preserving FL in Vehicular Ad-hoc Networks (VANETs). To address this, we propose EPDA, a dual aggregation scheme that enhances both efficiency and privacy. EPDA aggregates signatures for rapid verification and batches model updates to complete aggregation in a single round. It preserves model confidentiality and ensures anonymity and unlinkability. Experiments on the GTSRB dataset show that EPDA maintains low cryptographic cost (0.03s per client) and minimizes communication overhead to 148 bytes per client, scaling effectively to 350 clients. These results demonstrate EPDA’s practicality for secure and efficient FL in real-world VANETs.
Read the paper
Related studies
- Active Learning during Federated Learning for Object Detection
- Fairness-aware federated learning
- Decentralized Federated Learning
- Building a Personally Identifiable Information Recognizer in a Privacy Preserved Manner Using Automated Annotation and Federated Learning
- Preservation of Patient Level Privacy: Federated Classification and Calibration Models
- Efficient explainable federated learning
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.