In-vehicle network monitoring is one of the important elements in vehicular network management and security. Most of the existing network monitoring approaches rely on measuring every part of the network. Such approaches overburden the network by transmitting active probes. In this work, we propose a new in-vehicle network monitoring approach that benefits from network tomography and the advances in deep learning to infer the network delay performance. Specifically, the available measurements can be used to estimate the performance of the remaining network where direct measurements cannot be applied. Performance evaluation has been conducted using invehicle network simulation with different TSN (Time-Sensitive Network) traffics and the proposed monitoring approach shows the delay estimation accuracy of up to 99%.

No abstract

Due to the increased complexity of new in-vehicle networking architectures, which makes direct monitoring of internal network components intractable, alternative solutions are required to tackle this issue. One solution is to leverage the end-to-end measurements to estimate the internal network performance. To this end, we propose to employ network tomography as a monitoring tool for in-vehicle networks. Network tomography can infer the overall network performance by measuring only subset of the network. We investigate the use of network tomography in in-vehicle network by analysing network identifiability of three main architectures: bus-based, central-gateway, and Ethernet-based architectures. Our analysis results indicate the applicability of network tomography in in-vehicle networks based on certain topological and monitors' conditions. Furthermore, we validate our analytical results through simulation which shows a maximum error of only 0.174 milliseconds. Moreover, we compare the proposed approach with one of existing solutions and show that network tomography achieves better performance with minimal monitoring overhead of up to 52.2% and 782.3µs bandwidth and latency improvements, respectively.

With the advancements in Internet-of-Things (IoTs), particularly in Internet-of-Vehicles (IoVs), the vehicle becomes more vulnerable to more attack types caused by connecting the vehicle to the outside world. Moreover, the shift towards automotive Ethernet exposes the vehicle to IP-based attacks similar to attacks on computer networks. Most of such attacks tamper with the internal network components in order to gain control or disable some (or all) of the vehicle's functions. To this end, in this work, we study two in-vehicle network monitoring approaches based on network tomography. The first approach relies purely on deep neural network (DNN) and we call it DNN-based tomography approach, while the second was proposed in a previous work and it uses algebraic network tomography with deep neural network, we call this one DNN-based algebraic approach. We evaluated the inference performance of both approaches using simulations and found that the DNNbased algebraic tomography approach outperforms DNN-based tomography approach with less inference error of about 4.5µs.