mozilla/fxa-auth-db-mysql

Use base32 for recovery code generation, to avoid profanities and bias

Closed this issue · 4 comments

rfk commented

The algorithm we use for generating recovery codes does so by randomly generating 10 bytes, then mapping each byte onto one of the 36 alphanumeric characters:

https://github.com/mozilla/fxa-auth-db-mysql/blob/10bb799/lib/db/util.js#L114

The way this mapping is done has a slight statistical bias, since 36 does not evenly divide into 256. It also has the possibility of producing accidental profanity. Let's consider using the base32 approach that auth-server uses for unblock codes, which is unbiased and which omits the vowels most likely to produce the worst English-language profanities:

https://github.com/mozilla/fxa-auth-server/blob/master/lib/crypto/base32.js

vladikoff commented

@vbudhram I'll get @deeptibaghel to look at this

vbudhram commented

Hey @deeptibaghel thanks for taking this on. A couple of notes to help you get started

  • I am in favor of removing the configuration value that defines the recovery code keyspace. I don't believe it adds as much value as I intended.
  • Add a function to generate random base32 in utils, similar to auth-server base32. Since it is being used in once place, I think it would be ok to keep in utils and not be pulled into its own file.
  • Update generateRecoveryCodes to use base32 and create count number of codes

There will be some amount of tests that also need fixing. Let me know if you have any issues!

  • 👍1