VeCure: A practical security framework to protect the CAN bus of vehicles

Abstract

Vehicles are being revolutionized by integrating modern computing and communication technologies in order to improve both user experience and driving safety. As a result, vehicular systems that used to be closed systems are opening up various interfaces, such as Bluetooth, 3G/4G, GPS, etc., to the outside world, thus introducing new opportunities for cyber attacks. It has been recently demonstrated that modern vehicles are vulnerable to several remote attacks launched through Bluetooth and cellular interfaces, allowing the attacker to take full control of the vehicle. The common root cause of these attacks is the lack of message authentication for the vehicle's internal bus system, called Controller Area Network (CAN). In this work, we propose VeCure -a practical security framework for vehicular systems, which can fundamentally solve the message authentication issue of the CAN bus. VeCure is designed to be compatible with existing vehicle system architectures, and employs a trust group structure and a novel message authentication scheme with offline computation capability to minimize online message processing delay and deployment cost. We built a proof-of-concept prototype on a testbed using Freescale's automotive development boards. The experimental results show that VeCure only introduces 50us additional delay to process a message, which is at least 20-fold faster than any existing solution.

A Secure Communication Framework for ECUs

“…Malicious device records these identifiers and can launch attacks against other nodes. In [7], an authentication and encryption mechanism is described, but it requires about three times more CPU cycles than the cycles required for normal operation of a CAN node.…”

Evaluation of CAN Bus Security Challenges

“…VeCure [48] authentication, which has an acceptable 50 us processing delay, is based on trust groups where high-trust groups share a symmetric secret key. The method has a major advantage with fewer key numbers, which corresponds in size to the number of trust groups rather than the ECU number.…”

CopyCAN

“…Regarding authentication protocols, the general idea relies on using part of the CAN packet to transmit a hash of the message, encrypted through a secret key, alongside a counter (to defend from replay attacks). Although some state-of-the-art proposals like LeiA [24], VatiCAN [3] and VeCure [33] CAN, their downsides are still relevant: the decreases in response times and data bandwidth, generated respectively by the calculation and the transmission of the hash, make the usage of authentication protocols an unfeasible solution in most cases. In fact, their implementation is hardly feasible on less powerful ECUs, such as sensors, due to the computation requirements, while busy networks with many ECUs would suffer from the bandwidth overheads.…”