• Roots
• TechScript

Intro

Collection of technical notes for support of the Roots/NewRootsTaskForce and the Roots/CreationCeremony. Also see the SecurityManual.

Creating the Key with OpenSSL

Use these:

For research....

Random Feed

• openssl random number seed check: openssl provides a mechanism to look for EGD daemon socket eg /dev/entropy or from environment RANDFILE variable setting for file with random number. Use -rand option of openssl for random number seed (fifo) socket on private key generation.

Sources of RNGs (daemons) for mixing

• HAVEGE needs linux => takes randomness from internal states of the CPU which are more or less random as there are frequent system context switches

• /dev/random, which can be seeded via super user access and /dev/[us]random.

• turbid which provides mechanism for obtaining entropy via alsa audio Linux drivers (academic report). Must be calibrated. Turbid can seed OpenSSL via FIO or /dev/random. Use standard Turbid code. The paper describes the theoretical backgrounds and how to.

• ccd noise method.
  • One suggestion is photos of white or grey backgounds

  • another is here: LavaRnd

  • then hash the frames

Methods of mixing the RNs

• feed all sources all into /dev/random. Problems:

  • how do we know it is working? Use verbalized mode of program to show this is working. What is that?

  • can do these to check functionality:

cd /proc/sys/kernel/random/
cat uuid            # fetches random from the pool...
cat entropy_avail   # reveals how much is in the pool

• XOR all the different sources together. Problems:

  • need to read the code. E.g., mix.c .

  • (Previous version was buggy and did logical OR instead of XOR; I just replaced it. –NicoR)

  • Check the XOR method.

  • Thanks NicoR. You did read it. But, how are we going to know that others are reading the source code??? - Iang

• encrypt one feed with the key in the other feed. Problems:
  • need to read the code, and
  • need to chain it for more then 2 feeds.
• concatenate all sources and SHA2 them

  • it was pointed out on crypto that a better way would be to append all the different sources and hash the result, as XOR can be attacked if the other inputs can be seen

• concatenate by mixing on byte level

Tooling

OpenSSL

• http://www.openssl.org/source/ to download all versions (use FTP links to see hashes)

• ${WGET} -v -O ${OpenSSL}.tar.gz http://www.openssl.org/source/${OpenSSL}.tar.gz

• ${WGET} -v -O ${OpenSSL}.tar.gz.md5 http://www.openssl.org/source/${OpenSSL}.tar.gz.md5

OpenPGP

• GNUPG Integredity checking:

  • GNUPG=gnupg-1.4.9 GNUPG_SHA1="826f4bef1effce61c3799c8f7d3cc8313b340b55" GNU_MD5="cc52393087480ac8d245625004a6a30c"
  • GNUPG2=gnupg-2.0.9 GNUPG2_SHA1="959bdb934e3a72d256bfbd0122d996a73adb5d1f"

• ${WGET} -v -O ${GNUPG}.tar.bz2 ftp://ftp.gnupg.org/gcrypt/gnupg/${GNUPG}.tar.bz2

• ${WGET} -v -O ${GNUPG2}.tar.bz2 ftp://ftp.gnupg.org/gcrypt/gnupg/${GNUPG2}.tar.bz2

• echo "$GNUPG_MD5 ${GNUPG}.tar.bz2" >${GNUPG}.tar.bz2.md5

Random Numbers

• Turbid software with hardware tooling (audio cabling, Voltmeter) to generate seed with reported entropy (see study Turbid).
  • turbid_MD5="b5170bccaf06ec905a75042cc95ce965"

  • ${WGET} -v -O turbid.tgz http://www.av8n.com/turbid/turbid.tgz

  • turbid makes use of alsa audio drivers and sound card:
    • soundcard to be used: Creative SB PC128 CT4700, in full duplex mode
    • Turbid QA box, stereo audio cabling, Fluke 10 multimeter for saturation calibration
    • ENS1370 Soundcard specific turbid settings: 208mV saturation, Q=1.17e-05, Rin=25719 (card input impedance, Kout=1.47961, Bandwidth 4556, Vin/Vmeter 0.9996 (should be very close to 1).
    • There is a script available for redoing soundcard calibration on sound cards (estimated cal. time: 3 hours)

• webcam for cheap & cheerful piccy_rng

  • frame grabber bearshand_Ver.0.7.4.0e.zip BEARSHAND_MD5="ca0b974adbc496d96a7981c9697d8572"

  • http://moments-studio.com/en/BearsHand/

• /dev/random
  • must not use /dev/urandom

  • on Linux only, not on *BSD. On *BSD, /dev/random is synonymous with /dev/urandom . It is broken for entropy collection

Misc about /dev/random

Data we could put into /dev/random has to be put more or less slowly

So we could have a script taking bytes from several files, let's say 10 byte from each file at a time, to feed some randomness into the pool

so we can have files with some KB of random stuff from - a camera - turbid (with will be fed automatically) - a laptop - some havege data

The thing is to put the data as slowly as possible not to fill up the pool and loose the random data.

more on /dev/random in OpenSSLHowTo

We can look at /proc/sys/kernel/random/entropy_avail to see if data is feed by turbid or random sources files into the /dev/random pool

Hardware

• Stand alone Linux machine with standard CD live to generate private keys
• usb sticks for archiving passwords and keys
• video recorder for recording process.

Script

• CategoryAudit

• CategoryNewRootsTaskForce