[Cryptech Tech] Fast(er) warm up of the TRNG.

[Fredrik Thulin]

On Thursday 13 March 2014 17.29.34 Joachim Strömbergson wrote:
> Aloha!
> 
> Here is an idea:
> 
> How about if we provide unused seed values from previous operations as
> initial entropy? The idea is similar to what several OS:es do to shorten
> the time to operational /dev/random. In our case we would (for example)
> save a few seed (2-3 x 512 bits) in an FPGA-external battery powered
> SRAM, the same as we use for storing master key(s).
> 
> When the Cryptech HW-system is powering up, it check if there are any
> seed values stored. If there are, these seeds are read back from the
> SRAM into the TRNG and fed into the mixer part of the collector and used
> to create an initial seed. This would allow the TRNG to be able to start
> very fast (usec- msec) for all but the very first power up.

Sure. I'm thinking maybe we can model that as an entropy provider that 
provides entropy generate earlier? Maybe that works, maybe not. Depends on how 
we end up mixing entropy from providers of different speeds.

The previously generated entropy could be stored in an AEAD, inheriting the 
security from the tamper protection of the device.

/Fredrik