[hkhall]

I lived with the author for several years so I am coming at the paper from a position of knowing how he approached the work. I also see it as typical of systems and EE/CS MEng work at MIT.

In my grad program a Dr. Jason Dahlstrom [1] just graduated and his work with FPGAs to provide a limited attack vector surface is some of the more interesting FPGA work I have seen. The 10k' overview is you put some of the core features of the operating system inside FPGA hardware and define a controlled interface to it. You leverage the mass parallelism of the FPGA to have multiple instances running simultaneously and use a consensus protocol for output. Further you enable kill of each of these processes in a pseudorandom fashion and restarts from PROM so in case any one of them gets corrupted it won't have an effect for a long. I can't speak to all of the details but that is what I gleaned from a couple talks I attended in the past 2-3 years. Cool stuff.

[1] https://engineering.dartmouth.edu/people/faculty/jason-dahls...

edit: a space