Abstract

How well are the tools of modern cryptography employed in fielded embedded systems? How are the common tasks of communication and authentication, key storage and distribution, and firmware update and verification performed? In this talk, we describe evidence gathered from several studies of deployed embedded systems: a modern mass-market automobile and two electronic voting machines. These studies consisted of substantial reverse-engineering efforts by large teams of researchers. We find that in many cases the designers of the systems we studied are getting simple cryptographic tasks wrong. These failures suggest a lack of engagement with the cryptography and security research community. We consider some reasons for the status quo, and some ways that it might be improved. Joint work with Danny Anderson, Stephen Checkoway, Alexei Czeskis, Ariel Feldman, Edward Felten, J. Alex Halderman, Srinivas Inguva, Brian Kantor, Tadayoshi Kohno, Karl Koscher, Damon McCoy, Shwetak Patel, Eric Rescorla, Franziska Roesner, Stefan Savage, and Dan Wallach.