Improving in-vehicle network architectures using automated partitioning algorithms

Today´s in-vehicle networks are divided into domains using “best engineering practice”. However, as far as we are aware of, there are no existing tools that do this domain partitioning in an automated and optimal way. A strategy for designing in-vehicle networks is to group Electronic Control Units (ECUs) into domains so that each domain isolates a certain functionality and minimizes dependencies to other domains. In this paper, we use an automated partitioning algorithm and apply it to an in-vehicle network from a real, modern car, and we analyze the results from such an approach and compare it with the EVITA reference architecture. Different partitioning criteria can be used, and we investigate security domains based on both message types and on domains optimized to minimize inter-domain traffic. We show that our approach is very flexible and can identify meaningful in-vehicle network domains which are better than the EVITA domains with respect to communication, safety and security. We have also investigated the relationship between safety and security to see if security domains contradict or support partitions based on ASILs.