Secure autonomous cyber-physical systems through verifiable information flow control

Jed Liu, Joe Corbett-Davies, Andrew Ferraiuolo, Alexander Ivanov, Mulong Luo, G. Edward Suh, Andrew C. Myers, Mark Campbell
[doi] [Google Scholar] [DBLP] [Citeseer]
Read: 06 October 2019

Proceedings of the 2018 Workshop on Cyber-Physical Systems Security and Privacy,
Pages 48-59
Topic(s): security
Note(s): hardware, information flow

This paper ties together several different threads to create a CPS system (autonomous vehicle) that is resistant to many forms of attack. Builds on Jif (Java with security labels), SecVerilog (HDL with security labels), Hyperflow (processor with security labels) and statistical detection of attack based on sensor fusion (sensors plus map data). The Hyperflow processor is not used in the experiment at this stage because they need to port Java/Jif to that architecture. Statistics try to distinguish noise (Gaussian) from attack (uniform) — not clear to me that this is sufficiently robust.