Cascading dynamic slowdowns around hacked vehicles

Almost 40% of vehicles in the US can connect to the internet. While there are significant benefits of increased connectivity, hackers have illustrated that internet connected vehicles can be compromised remotely. Further, large-scale hacks have the potential to cause disastrous city-wide disruptions. Recently, it was shown using percolation theory that randomly stalling 20% of cars in Manhattan would cause a total freeze of city traffic. While this upper-bound estimate served as the first quantification of city-scale disruptions when vehicles are hacked, dynamic effects potentially cause significant slowdowns at lower fractions of hacked vehicles. Here, we perform simulations of traffic dynamics around hacked vehicles on grids of one-lane roads, using the SUMO traffic simulation platform. We consider slowdowns, when vehicles encounter road blocks from hacked vehicles and find that shortly post-hack, blocked roads cascade to neighboring regions, leading to significant slowdowns across the grid. We find that slowdowns are a function of hacked vehicle density, time since hack, and grid size. At large enough wait times, even low fractions of hacked vehicles cause significant gridlocks. Our results provide insights for recovery and rerouting to mitigate impacts of a large-scale hack.