[necovek]

I wonder how common are setups where an internal person has access to the TLS private key part of the certificate or access to a network equipment that all traffic passes through, yet they cannot access the inputs required for hashing/encryption client-side?

This seems to mostly prevent accidental logging and is thus a matter of defense in depth, stopping malicious actors from exploiting it later — but an actively malicious IT person would not be deterred.

[jmull]

> This seems to mostly prevent accidental logging

Yes, and that's not uncommon, IME. There's generally a lot of logging that's at least potentially available, and it gets turned on, and the logs shared when there's a problem that needs to be fixed (especially when it needs to be fixed quickly, which is usual).

This is going to make more sense for "enterprise"-type deployments, where there's a significant distinction between the people who might have access to request logs at times, and the people who can push code to production.

[fc417fc802]

Yes limited protection against insiders is good defense in depth but not the primary purpose which is to protect end user accounts on other services in the event that you are breached.

[necovek]

My question still stands: how do you disallow cleartext password extraction if you are breached, assuming all your IT infrastructure and code is now accessible to an attacker?

I am talking about not logging them ever, using internal TLS and strong hashing in general, and wondering what exact value is added on top with client side hashing.

[fc417fc802]

There are substantial differences between database access, snooping the logs, internal (no TLS) wiretap, and full MITM of the frontend.

Hashing client side minimizes the risk of any blast radius exceeding the bounds of your own service. There's obviously no way to prevent an adversary who achieves full MITM from gradually harvesting credentials over time. The only solution there is to use keys instead of passwords.

[necovek]

We are not disagreeing, but I am not getting my answer: how is client side hashing really helping, what are the circumstances it helps with if you do have the basics right?

In your enumeration, what is breached for this to be meaningfully impactful for other services where customers might be reusing credentials?

[fc417fc802]

As opposed to what? Your question seems unclear to me.

I already answered you that if you assume full MITM of the frontend then it is physically impossible to prevent gradual credential harvesting. Did you have a different scenario in mind?

> how is client side hashing really helping

Compared to what? Server side hashing? It prevents the plaintext from ever hitting your infra which minimizes to the greatest extent possible an insider or intruder gaining access to the plaintext. Since preventing access to the plaintext is the primary purpose of hashing it's the sensible option if you're forced to pick only one.

Of course there's really no good reason to pick only one of the two. You might as well elect for defense in depth against rogue insiders with full db access even though it's difficult to imagine what use they would have for a password based login at that point.

[necovek]

There is certainly good reason to do server-side hashing: you do not keep a persistent record of the customer's secret, yet keep the ability for them to authenticate with it.

If you are never logging a clear-text secret and storing a hashes version to validate against, and using TLS between client and server, client-side hashing does not bring much benefit other than protecting customers' reused passwords against people who have sufficient access to the infrastructure to MITM the client but no ability to modify the client side code where they could extract the password directly.

When IT has write access to code, client side hashing protects only against accidental log leakage and similar.