[Cryptography] Photon beam splitters for "true" random number generation ?

Henry Baker hbaker1 at pipeline.com
Thu Dec 24 14:39:39 EST 2015


At 07:47 AM 12/24/2015, Patrick Chkoreff wrote:
>Maybe to do a full cleanse I could run this until it exhausts the file space: $ cat /dev/zero >giant
>
>It has to put the zeros *somewhere*, after all.

In the case of /dev/zero, actually no!

There's more than enough room in the flash memory controller chip firmware to do some sort of compression -- e.g., simple run length compression -- thus leaving plenty of space for keeping your secrets unmolested (& unerased !).

So your only hope is (see below)

cat /dev/random > /dev/sdc

(where "sdc" is the USB stick and/or SD card).

This is probably a pretty good test of a new chip anyway, just to make sure that every sector works, and that it isn't hiding malware -- at least in the user-accessible portions of the flash disk.

This just in.  I just ran some benchmarks on my little ol' Linux computer.

dd if=/dev/urandom of=/dev/sdc
runs at about 1 MB/sec, whereas

dd if=local_file of=/dev/sdc
runs at about 2.2 MB/sec for this slow USB flash drive (USB flash drive write speed limited)

dd if=/dev/random of=/dev/sdc
takes forever!

./infnoise > /dev/sdc
runs at about 1 MB/minute.

"infnoise" being the HW "infinite noise" TRNG.

Bottom line: if you want to "cleanse" your USB flash fast, write a huge, but legal (!), long movie to it; H.264 is pretty darn compressed; use H.265 if you want a slightly more compressed bit source -- e.g., so-called "4K" (resolution) movie.



More information about the cryptography mailing list