Argon2 - Ruby Wrapper

A ruby wrapper for the Argon2 password hashing algorithm.

This is an independent project, and not official from the PHC team.

This gem provides a 1:1 replacement for the argon2 gem, with various improvements. Want to know more about why argon2 was forked? Read more

Wish to upgrade an existing application to use the improved API? Migration guide

This fork is kept up-to-date with argon2, latest sync: argon2 - v2.2.0

1. Useful Links
2. API Summary
3. Installation
4. Why fork argon2?
5. Migrating from argon2 to sorcery-argon2
6. Contributing
7. Contact
8. License

• Documentation

Below is a summary of the library methods. Most method names are self explaining and the rest are commented:

# Class methods
Argon2::Password.create(password, options = {})
Argon2::Password.valid_hash?(digest)
Argon2::Password.verify_password(password, digest, pepper = nil)

# Instance Methods
argon2 = Argon2::Password.new(digest)
argon2 == other_argon2
argon2.matches?(password, pepper = nil)
argon2.to_s   # Returns the digest as a String
argon2.to_str # Also returns the digest as a String

# Argon2::Password Attributes (readonly)
argon2.digest
argon2.variant
argon2.version
argon2.t_cost
argon2.m_cost
argon2.p_cost
argon2.salt
argon2.checksum

Argon2::Error
Argon2::Errors::InvalidHash
Argon2::Errors::InvalidVersion
Argon2::Errors::InvalidCost
Argon2::Errors::InvalidTCost
Argon2::Errors::InvalidMCost
Argon2::Errors::InvalidPCost
Argon2::Errors::InvalidPassword
Argon2::Errors::InvalidSaltSize
Argon2::Errors::InvalidOutputLength
Argon2::Errors::ExtError

Add this line to your application's Gemfile:

gem 'sorcery-argon2'

And then execute:

$ bundle

Or install it yourself as:

$ gem install sorcery-argon2

Require Sorcery-Argon2 in your project:

require 'argon2'

While implementing Argon2 support in Sorcery v1, I noticed that the current ruby wrapper (argon2 - technion/ruby-argon2) had some questionable design decisions, and attempted to address them through a pull request. The sole maintainer of the gem rejected these changes summarily, without pointing out any specific concerns other than not understanding why the changes were necessary. This lead to me (@joshbuker) being directed to create a fork instead: technion/ruby-argon2#44

You shouldn't trust this code more than you trust any other open source project. It's written by someone you don't know, and even if there is no malicious intent, there is no guarantee that the code is secure. Open source security is driven by having the community vett popular libraries, and discovering flaws through the sheer number of intelligent community members looking at the code.

That being said, the original library argon2 also falls under the same category. Ultimately, it was also written by a single person and is not thoroughly vetted by the community at the time of writing. A community member (@joshbuker, in this case) finding flaws in the implementation, and the fixes being rejected from upstream, is how this fork came into being.

The Argon2::Password interface was, to put it bluntly, poorly executed in the original library. The Password class instance was not a representation of an Argon2 password as one would expect, but instead an unnecessary abstraction layer used to store the settings passed to the underlying Argon2 C Library. This not only led to an overly complicated method of generating Argon2 hashes, but also meant that the class could not be used to read data back out of an Argon2 digest.

Originally, to generate an Argon2 hash/digest, one would have to do the following:

# Create an instance of the Argon2::Password class to store your options:
instance = Argon2::Password.new(t_cost: 4, m_cost: 16)
# Use this instance to generate the hash by calling create:
instance.create(password)
    => "$argon2i$v=19$m=65536,t=2,p=1$jL7lLEAjDN+pY2cG1N8D2g$iwj1ueduCvm6B9YVjBSnAHu+6mKzqGmDW745ALR38Uo"

Not only is this abstraction step unnecessary, it opens up a new way for developers to make a security mistake. New salts are only generated on the creation of a new Argon2::Password instance, meaning if you reuse the instance, those passwords will share the same salt.

instance = Argon2::Password.new(t_cost: 4, m_cost: 16)
# digest1 and digest2 will share the same salt:
digest1 = instance.create(password1)
digest2 = instance.create(password2)

Also, because of how the instance of Argon2::Password was designed, it cannot be used for reading information back out of an Argon2::Password. This is a summary of the original Argon2::Password API:

# Class methods
Argon2::Password.create(password) # Uses the default options to create a digest
Argon2::Password.valid_hash?(digest)
Argon2::Password.verify_password(password, digest, pepper = nil)

# Instance Methods
argon2 = Argon2::Password.new(options = {}) # Purely for storing options
argon2.create(password) # Take the options and generate an Argon2 digest

Compare this with sorcery-argon2:

# Class methods
Argon2::Password.create(password, options = {}) # Same as before but accepts passing options
Argon2::Password.valid_hash?(digest)
Argon2::Password.verify_password(password, digest, pepper = nil)

# Instance Methods
argon2 = Argon2::Password.new(digest) # Now represents an Argon2 digest
argon2 == other_argon2 # Which can be compared with `==` against other Argon2::Password instances
argon2.matches?(password, pepper = nil) # Or against the original password
argon2.to_s   # Returns the digest as a String
argon2.to_str # Also returns the digest as a String

# Argon2::Password Attributes (readonly)
argon2.digest
argon2.variant
argon2.version
argon2.t_cost
argon2.m_cost
argon2.p_cost
argon2.salt
argon2.checksum

Another minor issue is that all library errors fall to a single non-descriptive class:

Argon2::ArgonHashFail

Compare with sorcery-argon2:

Argon2::Error # Replaces `Argon2::ArgonHashFail`

# The following errors all inherit from Argon2::Error, and allow you to catch
# specifically the error you're interested in:
Argon2::Errors::InvalidHash
Argon2::Errors::InvalidVersion
Argon2::Errors::InvalidCost
Argon2::Errors::InvalidTCost
Argon2::Errors::InvalidMCost
Argon2::Errors::InvalidPCost
Argon2::Errors::InvalidPassword
Argon2::Errors::InvalidSaltSize
Argon2::Errors::InvalidOutputLength
Argon2::Errors::ExtError

Finally, the original library documentation is not only incomplete, but straight up broken/inaccurate in some areas. sorcery-argon2 has fixed these issues, and has 100% documentation of the API.

• argon2 Documentation
• sorcery-argon2 Documentation

There are two primary changes going from argon2 to sorcery-argon2:

Argon2::Password.new and Argon2::Password.create are now different.

Argon2::Passwords can now be created without initializing an instance first.

To upgrade:

# Take instances where you abstract creating the password by first exposing an
# Object instance:
instance = Argon2::Password.new(m_cost: some_m_cost)
instance.create(input_password)

# And remove the abstraction step:
Argon2::Password.create(input_password, m_cost: some_m_cost)

Argon2::Password.create no longer accepts custom salts.

You should not be providing your own salt to the Argon2 algorithm (this library does it for you). Previously you could pass an option of salt_do_not_supply, which has been removed in sorcery-argon2 - v1.0.0.

The root level error has been renamed.

Argon2::ArgonHashFail has been renamed to Argon2::Error

To upgrade:

# Find any instances of Argon2::ArgonHashFail, for example...
def login(username, password)
  [...]
rescue Argon2::ArgonHashFail
  [...]
end

# And do a straight 1:1 replacement
def login(username, password)
  [...]
rescue Argon2::Error
  [...]
end

Bug reports and pull requests are welcome on GitHub at Sorcery/argon2.

Feel free to ask questions using these contact details:

Current Maintainers:

• Josh Buker (@joshbuker) | Email

This gem is available as open source under the terms of the MIT License.