A Multilayered Security Infrastructure for Connected Vehicles -- First Lessons from the Field

Connected vehicles are vulnerable to manipulation and a broad attack surface can be used to intrude in-vehicle networks from anywhere on earth. In this work, we present an integrated security infrastructure comprising network protection, monitoring, incident management, and counteractions, which we...

Full description

Saved in:
Bibliographic Details
Published in:arXiv.org 2023-10
Main Authors: Häckel, Timo, Meyer, Philipp, Stahlbock, Lukas, Langer, Falk, Eckhardt, Sebastian A, Korf, Franz, Schmidt, Thomas C
Format: Article
Language:English
Subjects:
Access control
Anomalies
Automobiles
Controllers
Ethernet
Fleet management
Infrastructure
Prototypes
Security
Software-defined networking
Vehicles
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Connected vehicles are vulnerable to manipulation and a broad attack surface can be used to intrude in-vehicle networks from anywhere on earth. In this work, we present an integrated security infrastructure comprising network protection, monitoring, incident management, and counteractions, which we built into a prototype based on a production car. Our vehicle implements a Software-Defined Networking Ethernet backbone to restrict communication routes, network anomaly detection to make misbehavior evident, virtual controller functions to enable agile countermeasures, and an automotive cloud defense center to analyse and manage incidents on vehicle fleets. We present first measurements and lessons learned from operating the prototype: many network attacks can be prevented through software-defined access control in the backbone; anomaly detection can reliably detect misbehavior but needs to improve on false positive rate; controller virtualization needs tailored frameworks to meet in-car requirements; and cloud defence enables fleet management and advanced countermeasures. Our findings indicate attack mitigation times in the vehicle from 257 ms to 328 ms and from 2,168 ms to 2,713 ms traversing the cloud.
ISSN:2331-8422