Vault Rails
Vault is the official Rails plugin for interacting with Vault by HashiCorp.
If you're viewing this README from GitHub on the master
branch, know that it may contain unreleased features or
different APIs than the most recently released version. Please see the Git tag that corresponds to your version of the
Vault Rails plugin for the proper documentation.
Table of Contents
Quick Start
β₯ back to top
-
Add to your Gemfile:
gem "vault-rails", require: false
and then run the
bundle
command to install. -
Create an initializer:
require "vault/rails" Vault::Rails.configure do |vault| # Use Vault in transit mode for encrypting and decrypting data. If # disabled, vault-rails will encrypt data in-memory using a similar # algorithm to Vault. The in-memory store uses a predictable encryption # which is great for development and test, but should _never_ be used in # production. Default: ENV["VAULT_RAILS_ENABLED"]. vault.enabled = Rails.env.production? # The name of the application. All encrypted keys in Vault will be # prefixed with this application name. If you change the name of the # application, you will need to migrate the encrypted data to the new # key namespace. Default: ENV["VAULT_RAILS_APPLICATION"]. vault.application = "my_app" # The address of the Vault server. Default: ENV["VAULT_ADDR"]. vault.address = "https://vault.corp" # The token to communicate with the Vault server. # Default: ENV["VAULT_TOKEN"]. vault.token = "abcd1234" end
For more customization, such as custom SSL certificates, please see the Vault Ruby documentation.
-
Add Vault to the model you want to encrypt:
class Person < ActiveRecord::Base include Vault::EncryptedModel vault_attribute :ssn end
Each attribute you want to encrypt must have a corresponding
attribute_encrypted
column in the database. For the above example:class AddEncryptedSSNToPerson < ActiveRecord::Migration add_column :people, :ssn_encrypted, :string end
That is it! The plugin will transparently handle the encryption and decryption of secrets with Vault:
person = Person.new person.ssn = "123-45-6789" person.save #=> true person.ssn_encrypted #=> "vault:v0:EE3EV8P5hyo9h..."
- Note The unencrypted value will still be saved if the attribute referenced has a corresponding column. (i.e.
ssn
in the case above)
Advanced Configuration
β₯ back to top
The following section details some of the more advanced configuration options for vault-rails. As a general rule, you should try to use vault-rails without these options until absolutely necessary.
Specifying the encrypted column
By default, the name of the encrypted column is #{column}_encrypted
. This is customizable by setting the :encrypted_column
option when declaring the attribute:
vault_attribute :credit_card,
encrypted_column: :cc_encrypted
- Note Changing this value for an existing application will make existing values no longer decryptable!
- Note This value cannot be the same name as the vault attribute!
Specifying a custom key
By default, the name of the key in Vault is #{app}_#{table}_#{column}
. This is customizable by setting the :key
option when declaring the attribute:
vault_attribute :credit_card,
key: "pci-data"
- Note Changing this value for an existing application will make existing values no longer decryptable!
Specifying a context (key derivation)
Vault Transit supports key derivation, which allows the same key to be used for multiple purposes by deriving a new key based on a context value.
The context can be specified as a string, symbol, or proc. Symbols (an instance method on the model) and procs are called for each encryption or decryption request, and should return a string.
- Note Changing the context or context generator for an attribute will make existing values no longer decryptable!
String
With a string, all records will use the same context for this attribute:
vault_attribute :credit_card,
context: "user-cc"
Symbol
When using a symbol, a method will be called on the record to compute the context:
belongs_to :user
vault_attribute :credit_card,
context: :encryption_context
def encryption_context
"user_#{user.id}"
end
Proc
Given a proc, it will be called each time to compute the context:
belongs_to :user
vault_attribute :credit_card,
context: ->(record) { "user_#{record.user.id}" }
The proc must take a single argument for the record.
Specifying a default value
An attribute can specify a default value, which will be set on initialization (.new
) or after loading the value from the database. The default will be set if the value is nil
.
vault_attribute :access_level,
default: "readonly"
vault_attribute :metadata,
serialize: :json,
default: {}
Specifying a different Vault path
By default, the path to the transit backend in Vault is transit/
. This is customizable by setting the :path
option when declaring the attribute:
vault_attribute :credit_card,
path: "transport"
- Note Changing this value for an existing application will make existing values no longer decryptable!
Lazy attribute decryption
By default, vault-rails
will decrypt a recordβs encrypted attributes on that recordβs initialization. You can configure an encrypted model to decrypt attributes lazily, which will prevent communication with Vault until an encrypted attributeβs getter method is called, at which point all of the recordβs encrypted attributes will be decrypted. This is useful if you do not always need access to encrypted attributes. For example:
class Person < ActiveRecord::Base
include Vault::EncryptedModel
vault_lazy_decrypt!
vault_attribute :ssn
end
# Without vault_lazy_decrypt:
person = Person.find(id) # Vault communication happens here
person.ssn
# => "123-45-6789"
# With vault_lazy_decrypt:
person = Person.find(id)
person.ssn # Vault communication happens here
# => "123-45-6789"
Single, lazy attribute decryption
By default, vault-rails
will decrypt all encrypted attributes on that recordβs initialization on a class by class basis. You can configure an encrypted model to decrypt attributes lazily and and individually. This will prevent vault from loading all vault_attributes defined on a class the moment one attribute is requested.
class Person < ActiveRecord::Base
include Vault::EncryptedModel
vault_lazy_decrypt!
vault_single_decrypt!
vault_attribute :ssn
vault_attribute :email
end
# Without vault_single_decrypt:
person = Person.find(id) # Vault communication happens here
person.ssn # Vault communication happens here, fetches both ssn and email
# => "123-45-6789"
# With vault_single_decrypt:
person = Person.find(id)
person.ssn # Vault communication happens here, fetches only ssn
# => "123-45-6789"
person.email # Vault communication happens here, fetches only email
# => "[email protected]"
Serialization
By default, all values are assumed to be "text" fields in the database. Sometimes it is beneficial for your application to work with a more flexible data structure (such as a Hash or Array). Vault-rails can automatically serialize and deserialize these structures for you:
vault_attribute :details,
serialize: :json,
default: {}
It is recommended to set a default matching type that you're serializing.
- Note You can view the source for the exact serialization and deserialization options, but they are intentionally not customizable and cannot be used for a full object marshal/unmarshal.
Custom Serializers
For customized solutions, you can also pass a module to the :serializer
key. This module must have the following API:
module MySerializer
# @param [String, nil] raw
# @return [String, nil]
def self.encode(raw); end
# @param [String, nil] raw
# @return [String, nil]
def self.decode(raw); end
end
Your class must account for nil
and "empty" values if necessary. Then specify the class as the serializer:
vault_attribute :details,
serialize: MySerializer
- Note It is possible to encode and decode entire Ruby objects using a custom serializer. Please do not do that. You will have a bad time.
Custom encoding/decoding
If a custom serializer seems too heavy, you can declare an :encode
and :decode
proc when declaring the attribute. Both options must be given:
vault_attribute :address,
encode: ->(raw) { raw.to_s.upcase },
decode: ->(raw) { raw.to_s }
- Note Changing the algorithm for encoding/decoding for an existing application will probably make the application crash when attempting to retrieve existing values!
Caveats
β₯ back to top
Mounting/Creating Keys in Vault
The Vault Rails plugin does not automatically mount a backend. It is assumed the proper backend is mounted and accessible by the given token. You can mount a transit backend like this:
$ vault mount transit
If you are running Vault 0.2.0 or later, the Vault Rails plugin will automatically create keys in the transit backend if it has permission. Here is an example policy to grant permissions:
# Allow renewal of leases for secrets
path "sys/renew/*" {
policy = "write"
}
# Allow renewal of token leases
path "auth/token/renew/*" {
policy = "write"
}
path "transit/encrypt/myapp_*" {
policy = "write"
}
path "transit/decrypt/myapp_*" {
policy = "write"
}
Note that you will need to have an out-of-band process to renew your Vault token.
For lower versions of Vault, the Vault Rails plugin does not automatically create transit keys in Vault. Instead, you should create keys for each column you plan to encrypt using a different policy, out-of-band from the Rails application. For example:
$ vault write transit/keys/<key> create=1
Unless customized, the name of the key will always be:
<app>_<table>_<column>
So for the example above, the key would be:
my_app_people_ssn
Searching Encrypted Attributes
Because each column is uniquely encrypted, it is not possible to search for a
particular plain-text value. For example, if the ssn
attribute is encrypted,
the following will NOT work:
Person.where(ssn: "123-45-6789")
This is because the database is unaware of the plain-text data (which is part of the security model).
Development
β₯ back to top
- Clone the project on GitHub
- Create a feature branch
- Submit a Pull Request
Important Notes:
- All new features must include test coverage. At a bare minimum, Unit tests are required. It is preferred if you include acceptance tests as well.
- The tests must be be idempotent. The HTTP calls made during a test should be able to be run over and over.
- Tests are order independent. The default RSpec configuration randomizes the test order, so this should not be a problem.