Authenticating the Sender on CAN Bus using Inimitable Physical Characteristics of the Transmitter and Channel

Abstract

The Cybersecurity for the embedded systems has become a serious challenge in the recent times. Given that the embedded applications are being connected with each other and over the public internet while running the relatively fragile low-density code, they are prone to a wide range of attacks. These attack surfaces are inherent to most of the embedded applications. One such example is a modern automobile. A modern vehicle consists of a network of small electronic computers known as Electronic Control Units (ECUs), which makes possible the state-of-the art features. Because of the power of these tiny computers and the artificial intelligence, autonomous vehicles will be on the road for public use in near future. These vehicles will be connected over the internet and hence susceptible to the broad range of attacks. The problem gets worse in the automotive applications because of the presence of very weak internal networking protocols. The ECUs are connected via each other over Controller Area Network (CAN) Bus which lacks the basic security features. It does not provide the authenticity of the message sender and the payload integrity is absent as well. In this paper, we have proposed a novel idea to solve both of these problems based on the physical fingerprinting the transmitter of the message packet. Electrical devices are unique in terms of the physical fingerprints, they leave in the transmitted messages due to the material’s microstructure. This uniqueness exists in the time domain as well as the frequency domain of the signals. We have proposed various techniques to capture this uniqueness using the signal processing techniques at the message receiver side which will be able to link the received packet to the original transmitter. We have applied the Neural Network based Classifier in order to realize an Intrusion Detection System proof of concept. Our proposed idea, realized with different techniques, has been proven to be more efficient than the state-of-the art intrusion detection systems. We have analyzed the weaknesses in one of the advanced security techniques based on fingerprinting the clock behaviors of the message sender. We were able to launch the successful attack to bypass the intrusion detection system based on fingerprinting the clock behavior of the sender. Our work demonstrates the wide range of attacks: the external attacks by exploiting the in-vehicle infotainment system, internal attacks and a possible defense mechanism as well. We have summarized the possible attack vectors on our proposed idea as well with the challenges being faced for the real-world implementation.