Voltage Based Electronic Control Unit (ECU) Identification with Convolutional Neural Networks and Walsh–Hadamard Transform

This paper proposes an identification approach for the Electronic Control Units (ECUs) in vehicle, which are based on physical characteristics of ECUs extracted from their voltage output. Then, is not cryptographic means, but it could be used as alternative or complementary means to strengthen solutions vehicle cybersecurity. While previous research has hand-crafted features such mean voltage, max skew variance, this study applies Convolutional Neural Networks (CNNs) combination with Walsh–Hadamard Transform (WHT), useful properties compactness and robustness noise. These exploited by CNN, particular, pooling layers, reduce size feature maps CNN. The proposed applied a recently public data set ECU fingerprints different automotive vehicles. results show that CNN WHT outperforms, terms accuracy, noise computing times, other approaches literature shallow machine learning tailor-made features, well linear transforms Discrete Fourier (DFT) original time domain representations.