Thread: Reg: Alternate way of hashing database role passwords

"M, Anbazhagan" <Anbazhagan.M@netapp.com> writes:
> Currently we are using SHA-256 default for password_encryption in our postgresql deployments. Is there any active
workbeing done for adding additional hashing options like PBKDF2, HKDF, SCRYPT or Argon2 password hashing functions,
eitherof which is only accepted as a algorithms that should be used for encrypting or hashing the password at storage
asper the Organization's Cryptography Standard. 

> If it is not in current plan, is there a plan to include that in subsequent versions?

It is not, and I doubt we have any interest in dramatically expanding
the set of allowed password hashes.  Adding SCRAM was enough work and
created a lot of client-v-server and cross-version incompatibility
already; nobody is in a hurry to repeat that.  Moreover, I know of
no reason to think that SHA-256 isn't perfectly adequate.

            regards, tom lane

On Wed, Jun 26, 2024 at 12:11 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> It is not, and I doubt we have any interest in dramatically expanding
> the set of allowed password hashes.  Adding SCRAM was enough work and
> created a lot of client-v-server and cross-version incompatibility
> already; nobody is in a hurry to repeat that.  Moreover, I know of
> no reason to think that SHA-256 isn't perfectly adequate.

If history is any guide, every algorithm will eventually look too
weak. It seems inevitable that we're going to have to keep changing
algorithms as time passes. However, it seems like SCRAM is designed so
that different hash functions can be substituted into it, so what I'm
hoping is that we can keep SCRAM and just replace SCRAM-SHA-256 with
SCRAM-WHATEVER when SHA-256 starts to look too weak.

What I find a bit surprising about Anbazhagan's question is that he
asks about PBKDF2, which seems to be part of SCRAM already.[1] In
fact, I think all the things he lists are key derivation functions,
not hash functions. I'm far from a cryptography expert, but it seems
surprising to me that somebody would be concerned about the KDF rather
than the hash function. We know that people get concerned about code
that still uses MD5, for example, or SHA-1, but this is the first time
I can remember someone expressing a concern about a KDF.

Anbazhagan, or anyone, is there some reason to think that the PBKDF2
approach used by SCRAM is a problem?

--
Robert Haas
EDB: http://www.enterprisedb.com

[1]
https://en.wikipedia.org/wiki/Salted_Challenge_Response_Authentication_Mechanism#Password-based_derived_key,_or_salted_password

Robert Haas <robertmhaas@gmail.com> writes:
> On Wed, Jun 26, 2024 at 12:11 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> It is not, and I doubt we have any interest in dramatically expanding
>> the set of allowed password hashes.  Adding SCRAM was enough work and
>> created a lot of client-v-server and cross-version incompatibility
>> already; nobody is in a hurry to repeat that.  Moreover, I know of
>> no reason to think that SHA-256 isn't perfectly adequate.

> If history is any guide, every algorithm will eventually look too
> weak. It seems inevitable that we're going to have to keep changing
> algorithms as time passes. However, it seems like SCRAM is designed so
> that different hash functions can be substituted into it, so what I'm
> hoping is that we can keep SCRAM and just replace SCRAM-SHA-256 with
> SCRAM-WHATEVER when SHA-256 starts to look too weak.

Totally agreed, that day will come.  What I'm pushing back on is the
suggestion that we should implement a ton of variant password hash
functionality on the basis of somebody's whim.  The costs are large
and they are not all paid by us, so the bar to replacing any part
of that has to be very high.

> What I find a bit surprising about Anbazhagan's question is that he
> asks about PBKDF2, which seems to be part of SCRAM already.[1] In
> fact, I think all the things he lists are key derivation functions,
> not hash functions.

This I don't have any info about.

            regards, tom lane