update scrypt params based on a stack overflow comment. still need to audit!
This commit is contained in:
parent
a6d41df4e4
commit
5c5b7b794c
1 changed files with 15 additions and 4 deletions
|
@ -154,12 +154,23 @@ def derive_secrets(root_password, salt):
|
|||
# TODO - Audit me audit me audit me! I don't know if these values are
|
||||
# optimal.
|
||||
#
|
||||
# I will say that it seems like there's an optimal for access control, and
|
||||
# there's a stronger optimal for sensitive storage.
|
||||
# TODO - try hashlib.scrypt? see if the values are the same? And maybe
|
||||
# switch to it, that sounds less bad than "hazmat". Also just look it up
|
||||
# maybe there's an answer as to why both are around.
|
||||
#
|
||||
# TODO - wallet_id in the salt? (with domain etc if we go that way)
|
||||
# But, we probably want random salt anyway for each domain, who cares
|
||||
scrypt_n = 1<<13
|
||||
scrypt_r = 16
|
||||
#
|
||||
# TODO - save scrypt parameters with the keys so we can change parameters
|
||||
# and still read old keys?
|
||||
#
|
||||
# https://stackoverflow.com/a/12581268
|
||||
# Per this, there's an optimal for interactive use, and there's a stronger
|
||||
# optimal for sensitive storage. Going with the latter since we're storing
|
||||
# encrypted stuff on a server. That said, that's based on presentation
|
||||
# slides from 2009. Maybe I should go even more secure?
|
||||
scrypt_n = 1<<20
|
||||
scrypt_r = 8
|
||||
scrypt_p = 1
|
||||
|
||||
key_length = 32
|
||||
|
|
Loading…
Reference in a new issue