Revision as of 20:03, 25 June 2021

Use yescrypt as default hashing method for shadow passwords

Summary

Make the yescrypt hashing method the default method used for new user passwords stored in /etc/shadow.

Owner

Name: Björn Esser
Email: besser82@fedoraproject.org

Current status

Targeted release: Fedora Linux 35
Last updated: 2021-06-25
FESCo issue: #2623
Tracker bug: #1975402
Release notes tracker: #706

Detailed Description

yescrypt is a password-based key derivation function (KDF) and password hashing scheme. It builds upon Colin Percival's scrypt, and is based on NIST-approved primitives.

Cryptographic security of yescrypt (collision resistance, preimage and second preimage resistance) is based on that of SHA-256, HMAC, and PBKDF2. Even a catastrophic failure of yescrypt’s computational layers to maintain entropy would not affect yescrypt’s cryptographic properties as long as SHA-256, HMAC, and PBKDF2 remain unbroken. That said, in case SHA-256 is ever broken, yescrypt’s additional processing is likely to neutralize the effect of any such break.

By the time of this writing, sha256crypt and sha512crypt, as used commonly today for hashing passwords, remain unbroken, but have some flaws by design:

Both hashing methods effectively only use about 90 bits of salt, although the NIST-recommendation for salt length is >= 128 bits.
Long passwords can create a denial-of-service on the CPU.
Passive observation of execution times can predict password length.
No use of a crytographic key derivation function (KDF).

In conclusion we should move to a stronger hashing method for computing the entries in the UNIX shadow file. So why not Argon2?

yescrypt has a dependency not only on RAM, but also on fast on-die local memory, which provides bcrypt-like anti-GPU properties even at very low per-hash RAM sizes (where Argon2 might even lose to bcrypt in terms of GPU attack speed).
yescrypt currently has less low-level parallelism within processing of a block, yet allows for tuning it later, whereas Argon2 has a fixed and currently commonly excessive amount of such parallelism, which may be extracted to speed up e.g. GPU attacks through use of more computing resources per the same total memory size due to each hash computation's memory needs being split between 32 threads (yescrypt currently has four 16-byte lanes that can be processed in parallel within a 64-byte sub-block before running into a likely data dependency for the next sub-block, whereas Argon2 allows for parallel processing of eight 128-byte chunks within a 1 KiB block with only two synchronization points for the entire block, as well as of four 32-byte parts of the 128-byte chunks with only two more synchronization points for the entire 1 KiB block).
yescrypt's cryptographic security is provided by SHA-256, HMAC, and PBKDF2, which are NIST-approved and time-tested (the rest of yescrypt's processing, while most crucial for its offline attack resistance properties, provably does not affect its basic cryptographic hash properties), whereas Argon2 relies on the newer BLAKE2 (either choice is just fine for security, but use of approved algorithms may sometimes be required for compliance)

Also see yescrypt - scalable KDF and password hashing scheme, the PHC submission paper, PHC yescrypt vs. Argon2, and the discussion on the Debian bugtracker.

Feedback

Feedback so far:

2 times: +1 from users on the Fedora-devel-list.
Some people see possible problems, if one copies the shadow-file between different distributions and/or systems, and/or uses some distributed user/pw management service like LDAP.
It is agreed upon, we shouldn't force users to change their login password; rehashing the current password silently is no problem for them, though.

Benefit to Fedora

yescrypt is the default password hashing scheme on recent ALT Linux, Debian testing, and Kali Linux 2021.1+, so we should adopt it as the default, too. Also, it is already the recommended hashing method in the Fedora CoreOS documentation.

Scope

Proposal owners: Help with integration for yescrypt support in some packages. See Dependencies.
Other developers: Integrate yescrypt support in some packages. See Dependencies.
Release engineering: #10150
Policies and guidelines: N/A (not needed for this Change)
Trademark approval: N/A (not needed for this Change)
Alignment with Objectives: N/A (not needed for this Change)

Upgrade/compatibility impact

No impact, as password hashes, that have been computed using the former default sha512crypt, will continue to work.

How To Test

Existing installations: Change your user password and check whether the computed password hash in /etc/shadow starts with $y$ , like root:$y$j9T$JEFtZ/….
Fresh installations: Check whether the password hash(es) for the user(s) created by anaconda in /etc/shadow start(s) with $y$ , like root:$y$j9T$JEFtZ/….

User Experience

No user visible changes, but they can rely on safer hashing for their user passwords.

Dependencies

accountsservice: upstream: https://gitlab.freedesktop.org/accountsservice/accountsservice/-/merge_requests/74, downstream: https://src.fedoraproject.org/rpms/accountsservice/pull-request/4
authselect: upstream: https://github.com/authselect/authselect/pull/253, downstream patch is in Rawhide.
libuser: upstream: https://pagure.io/libuser/pull-request/49, downstream patch is in Rawhide.
pam: upstream: support for yescrypt is in v1.4.0 or later, downstream (custom config) patch is in Rawhide.

anaconda: yescrypt will be used as default for created password hashes with anaconda >= 35.18.
guestfs-tools: Support for yescrypt is in v1.47.2 or later.
libxcrypt: Is already capable for computing yescrypt hashes since v4.3.
shadow-utils: Will support yescrypt in the next release. We carry the patches downstream until then.
systemd: Defaults to yescrypt already.

Contingency Plan

Blocks release? Yes

Partially revert the changes, that have been applied to accountsservice, anaconda, authselect, libuser, pam, and shadow-utils, and rebuild those packages.

Documentation

Fedora now uses the yescrypt hashing method for new passwords. There are no visible changes nor impacts to the end-user. Users are recommended to change their existing passwords after upgrading.

Release Notes