Arkana
noun ‧ secret knowledge¹ — ar‧ka‧na | \är-ˈkā-nə\
Store your keys and secrets away from your source code. Designed for Android and iOS projects.
Requirements
iOS
Your project must be using Swift Package Manager or CocoaPods as dependency manager (or both). No support for Carthage.
Note: this gem was only tested in macOS environments.
Android
The Java/Kotlin code generator hasn't been implemented yet. Star and "watch" this project and check back in the future, or help us build it here.
Preview
Click here to show an image with the preview!
The image below shows how the auto-generated file looks like. At the bottom of it you can see how you'll consume the code generated.
Installation
Add this gem to your Gemfile
if you're using bundler (recommended):
gem 'arkana'
And then run bundle install
to install it.
Alternatively, you can install it in your entire system instead (not recommended):
gem install arkana
Usage
Arkana requires the declaration of a YAML config file. Although you can name it whatever, the convention is to name it .arkana.yml
. See template.yml for practical examples.
Once you have create your config file, you can run Arkana:
Usage: arkana [options]
-c /path/to/your/.arkana.yml, Path to your config file. Defaults to '.arkana.yml'
--config-filepath
-e /path/to/your/.env, Path to your dotenv file. Defaults to '.env' if one exists.
--dotenv-filepath
-f, --flavor FrostedFlakes Flavors are useful, for instance, when generating secrets for white-label projects. See the README for more information
Note that you have to prepend bundle exec
before arkana
if you manage your dependencies via bundler, as recommended.
Arkana only has one command, which parses your config file and env vars, generating all the code needed. Arkana should always be run before attempting to build your project, to make sure the files exist and are up-to-date (according to the current config file). This means you might need to add the Arkana run command in your CI/CD scripts, fastlane, Xcode Build Phases, or something similar.
Importing Arkana into your project
Once the Arkana has been run, its files will be created according to the package_manager
setting defined in your config file, so update that setting according to your project needs.
Via Swift Package Manager
If you're integrating Arkana via SPM (package_manager
is set to spm
), Arkana will generate its files as a local Swift Package.
You can add this package in an Xcode project or as a dependency of another Swift Package:
Adding a local Swift Package to your Xcode project
- Choose File → Add Packages… and click on Add Local…. Locate and select the
ArkanaKeys
folder (or the name of theimport_name
option that you passed in your config file). - Select your project in the Project navigator, then select your app target and navigate to its General pane.
- Click the + button in the
Frameworks, Libraries, and Embedded Content
section, select the local package’s library product, and add it as a dependency.
Adding a local Swift Package to another Swift Package
Add ArkanaKeys
(or the name of the import_name
option that you passed in your config file) to your list of dependencies in your Package.swift
file, like this:
…
dependencies: [
…
.package(name: "ArkanaKeys", path: "path/to/your/dependencies/ArkanaKeys"),
…
],
…
Via CocoaPods
If you're integrating Arkana via CocoaPods (package_manager
is set to cocoapods
), Arkana will generate its files as a Development Pod.
Add ArkanaKeys
and ArkanaKeysInterfaces
(or the pod_name
option that you passed in your config file) to your list of dependencies in your Podfile
file, like this:
pod "ArkanaKeys", path: "path/to/your/dependencies/ArkanaKeys"
pod "ArkanaKeysInterfaces", path: "path/to/your/dependencies/ArkanaKeysInterfaces"
After adding its dependency, you should be able to import ArkanaKeys
(or the import_name
option that you passed in your config file).
We recommend you to add your ArkanaKeys directory to your .gitignore
since it's an auto-generated code that will change every time you run Arkana (since its salt gets generated on each run). For more information, see How does it work?
Options
--help
Will display a list of the available options.
--config-file-path
Usage: --config-file-path /path/to/your/.arkana.yml
Indicates where your config file is located at.
Defaults to .arkana.yml
in the current directory.
--dotenv-file-path
Usage: --dotenv-file-path /path/to/your/.env
Indicates where your .env file is located at, if you have any to be loaded.
Defaults to .env
in the current directory.
--flavor
Usage: --flavor loremipsum
Passing a flavor option will change the env var lookup mechanism. Flavors are useful, for instance, when generating secrets for white-label projects.
Example
Let's load a flavor called snowflakes
and load a secret called MySecretAPIKey
:
Run bundle exec arkana --flavor snowflakes
This will change the env var lookup method, to this particular order:
- Load any
.env
file, if present. - Load any
.env.snowflakes
, if present (it will override the keys of.env
when they conflict) - Look up for a
SnowflakesMySecretAPIKey
env var. This could be present anywhere: both in the.env
file, or in actual env vars (useful for CI environments). - If
SnowflakesMySecretAPIKey
didn't exist, look up for aMySecretAPIKey
env var. - If ultimately it couldn't find any env var to populate the key
MySecretAPIKey
, it will throw a validation error.
This means that, if you are working with a white-label project, you can have all your env vars declared in a single .env
file, or in multiple files like .env.snowflakes
, .env.frosties
, etc. This also means that your CI can be configured with the appropriate env vars e.g. SnowflakesMySecretAPIKey
, FrostiesMySecretAPIKey
, etc, with no necessity to manage (or git-version) dotenv files at all, which is the ideal way to manage secrets securely in a project.
Advanced usage
Continuous Integration
We advise you not to commit your .env
files, because of security concerns. They should live in secure Environment Variables in your build (CI/CD) server instead.
Following the template.yml example file, these would be the variables that would need to be added to your build server env vars:
FrootLoopsAppStoreAppID
FrootLoopsBackendDomain
FrootLoopsMySecretAPIKey
FrootLoopsMyServiceAPIKeyDebug
FrootLoopsMyServiceAPIKeyRelease
FrostedFlakesAppStoreAppID
FrostedFlakesBackendDomain
FrostedFlakesMySecretAPIKey
FrostedFlakesMyServiceAPIKeyDebug
FrostedFlakesMyServiceAPIKeyRelease
It's okay to commit your .env
files if their potential exposure wouldn't be harmful to your project, for instance when Arkana is being used in a white label project to inject variables but that are not necessarily "secrets" (e.g. app tint color, server domain, etc). You can also use .env
files to store part of your env vars (only the unsecure ones), and keep the secrets in your build server's env vars.
Monorepo
If your repository makes use of a monorepo structure, then we recommend using Arkana by defining multiple .env
files and multiple .arkana.yml
configuration files, as it's how it was designed to be used.
One of the goals of Arkana was to use the least dotfiles as possible. This made a significant difference when designing for white-label projects (aka project flavors), because traditional dotenv implementation would suggest having multiple dotenv files such as .env.frootloops
and .env.frosted
if you had 2 different flavors (note that you can still use multiple dotenv files if you want so (see usage of the --flavor
option). But what if you're a big agency that distributes to 25, 50, 200 clients? Maybe you don't want to be managing the distribution of 200 dotenv files across your team, and Arkana was designed with that in mind.
However, when it came to monorepo structures, we weighted the pros and cons of both approaches, and ultimately decided to the simplicity of having a set of dotfiles for each individual project, in their respective directory. The reason main reasons were:
- the scale here is different: a project may have hundreds of flavors, but not that many projects.
- each project already has its own dotfiles in its root directory, most likely, including
.env
files, and with Arkana it wouldn't be any different. - each project has its own quirks, so it's more scalable to put the responsibility on the project's directory to deal with its own dotenv files than to have a centralized file that would coordinate each project.
- the complexity of maintaining a centralized configuration file that manages all projects would be too high to be implemented in Arkana. We aim for simplicity.
If you have questions on how to set this up, feel free to open an issue and we can clarify further how this can be set up.
Presence of special characters in your env vars
Dollar sign
This project is implemented in Ruby and uses the dotenv
gem. Since dotenv
follows bash implementation as close as possible, dollar signs ($
) need to be escaped unless they are in single quotes.
For example, these are all valid:
SecretWithDollarSignEscapedAndAndNoQuotesKey = real_\$lim_shady
SecretWithDollarSignEscapedAndDoubleQuoteKey = "real_\$lim_shady"
SecretWithDollarSignNotEscapedAndSingleQuoteKey = 'real_$lim_shady'
These are not valid:
SecretWithDollarSignNotEscapedAndDoubleQuotesKey = "real_$lim_shady"
SecretWithDollarSignNotEscapedAndNoQuotesKey = real_$lim_shady
When storing your secret in actual env vars (instead of dotfiles), you will most likely need to escape them too.
Other characters to avoid
\
aka backslash: double scaping will be needed"
aka double quotes: escaping is needed
Other common special characters that are fine to be used
` ~ ! @ # % ^ & * ( ) _ - + = { [ } } | : ; ' < , > . ? /
FAQ
How does it work?
Arkana uses code generation to provide your app with its secrets. Secrets are fetched from env vars during Arkana runtime (not your app's runtime), their values are encoded using a salt that is generated on each run, and source code is generated using the provided keys, and the generated encoded values.
During your app's runtime, the encoded value is decoded so your app can use their raw values (the values that were originally stored in your env vars).
This encoding mechanism makes it difficult for attackers to simply just read your secrets in plain text from your app's binary (for instance by using unix strings, or other tools like dumpdecrypted).
Is this safe?
Key security is difficult. Right now even the biggest apps get their keys leaked. This is neatly summed up by John Adams of the Twitter Security Team on Quora.
Putting this in the context of, "should you be storing keys in software", is more appropriate. Many companies do this. It's never a good idea.
When developers do that, other developers can use debuggers and string searching commands to extract those keys from the running application. There are numerous talks on how to do that, but leave that as an exercise to the reader to find those talks.
Many people believe that obfuscating these keys in code will help. It usually won't because you can just run a debugger and find the fully functional keys.
So in summary, the ideal way to store keys is to not store keys. In reality though most Apps embed keys, and this does that and adds some rudimentary obfuscation to the keys. A well motivated app cracker could probably extract this within a few minutes however.
This excerpt has been copied in its entirety from https://github.com/orta/cocoapods-keys#security - 100% credit goes to @orta and its maintainers.
cocoapods-keys
?
Why not I decided to create this new gem because cocoapods-keys
wasn't enough for all my (and other cocoapods-keys
users') needs. The key differences between these two projects are:
- The code generation process generates a more modern interface and protocols, under a friendly and customizable namespace, with unit tests written in the output language, and modularized architecture.
- Flexibility: it's independent from CocoaPods. It's a CLI tool, which may or may not be used alongside CocoaPods, but also supports Swift Package Manager.
- Extensibility: can be extended to generate secrets for Android projects as well as iOS ones.
- Conciseness: built-in support for white-label projects and environment-specific secrets, properly making use of interfaces, and reducing amount of boilerplate configuration code to be written.
Why not a CocoaPods Plugin?
Because plugins can only be hooked to Podfiles, and not Podspec files. Thus, if you have local CocoaPods and rely only on their podspec files to generate their content, but they are consuming the secrets, then you're out of luck. The strategy that Arkana uses requires a little bit more manual work (1 LOC), but it's more flexible, reaching more projects with different setups.
If your setup uses CocoaPods, you can add this snippet at the top of your Podfile that would work just like a CocoaPods Plugin:
`bundle exec arkana --config-filepath /path/to/your/.arkana.yml --dotenv-filepath /path/to/your/.env`
Note: include the back-ticks above, they're required, so that Ruby interprets the snippet as a shell script.
By doing this, your Arkana code generation will be executed everytime you run pod install
or pod update
.
The only down side of using this strategy is that you wouldn't want to use it if you need to pass in (dynamic) flavors, since it would require you to modify the Podfile on every different flavor you need to build. If that's the case, you should run Arkana before running pod install
in your build pipeline, and not use this snippet at all.
References
Did you know?
The word arcanum (pluralized as "arcana", here spelled as "Arkana") came from Latin arcanus, meaning "secret", and entered English as the Dark Ages gave way to the Renaissance.²
Special thanks to @danilobecke for the inspiration and heavy lifting, and to @orta for the creation of https://github.com/orta/cocoapods-keys which this project is based off of.
Development
After checking out the repo, run bin/setup
to install dependencies. Then, run rake spec
to run the tests. You can also run bin/console
for an interactive prompt that will allow you to experiment.
To install this gem onto your local machine, run bundle exec rake install
.
To bump the lib's version, run bundle exec rake bump[1.2.3]
(replacing the value with the desired version).
To release a new version, update the version number in version.rb
(likely done via rake bump
above), and then run bundle exec rake release
, which will create a git tag for the version, push git commits and the created tag, and push the .gem
file to rubygems.org.
Contributing
If you spot something wrong, missing, or if you'd like to propose improvements to this project, please open an Issue or a Pull Request with your ideas and I promise to get back to you within 24 hours! 😇
This project is intended to be a safe, welcoming space for collaboration, and contributors are expected to adhere to the code of conduct.
For a list of issues worth tackling check out: https://github.com/rogerluan/arkana/issues
Popularity
License
This project is open source and covered by a standard 2-clause BSD license. That means you can use (publicly, commercially and privately), modify and distribute this project's content, as long as you mention Roger Oba as the original author of this code and reproduce the LICENSE text inside your app, repository, project or research paper.
Contact
Twitter: @rogerluan_