A Secure and Privacy Preserved Infrastructure for VANETs based on Federated Learning with Local Differential Privacy
Journal article
Authors | Batool, H., Anjum, A., Khan, A., Izzo, S., Mazzocca, C. and Jeon, G. |
---|---|
Abstract | Advancements in Vehicular ad-hoc Network (VANET) technology have led to a growing network of interconnected devices, including edge devices, resulting in substantial data generation. The data generated by vehicles is subsequently shared with other devices, such as Roadside Units (RSUs). However, ensuring secure data sharing poses a significant challenge due to the potential risk of data breaches. Recently, Federated Learning (FL) has garnered substantial attention in the research community, enabling data owners to collaboratively learn a shared prediction model while retaining all their training data privately. However, traditional FL-based approaches are susceptible to inference and gradient leakage attacks. This paper presents a framework for private data sharing in VANETs using FL with local differential privacy. In the first layer, vehicles apply local differential privacy techniques to their data before sharing it with the RSU. The second layer is responsible for training model parameters at the RSU and updating the trained weights with the training server. To assess our system's performance, we evaluate it based on accuracy and simulation time for both local and global parameter sharing. Additionally, we measure each client's performance by calculating accuracy measures during each iteration. The experimental results demonstrate that our framework not only ensures security against inference and gradient leakage attacks but also exhibits superior efficiency compared to its counterparts. |
Keywords | Differential Privacy; VANETs; Federated Learning; Laplace mechanism; Local and Global model; Gradient Leakage |
Year | 2023 |
Journal | Elsevier Information Sciences |
Journal citation | 652 |
Publisher | Elseiver |
ISSN | 0020-0255 |
Digital Object Identifier (DOI) | https://doi.org/10.1016/j.ins.2023.119717 |
Web address (URL) | https://www.sciencedirect.com/science/article/pii/S0020025523013026?utm_campaign=STMJ_AUTH_SERV_PUBLISHED&utm_medium=email&utm_acid=265827134&SIS_ID=&dgcid=STMJ_AUTH_SERV_PUBLISHED&CMX_ID=&utm_in=DM417386&utm_source=AC_ |
Accepted author manuscript | License File Access Level Controlled |
Output status | Published |
Publication dates | |
Online | 10 Oct 2023 |
Publication process dates | |
Accepted | 21 Sep 2023 |
Deposited | 20 Oct 2023 |
https://repository.derby.ac.uk/item/q26x5/a-secure-and-privacy-preserved-infrastructure-for-vanets-based-on-federated-learning-with-local-differential-privacy
Restricted files
Accepted author manuscript
36
total views1
total downloads3
views this month0
downloads this month