Zen Rails Security Checklist

Summary

This document provides a not necessarily comprehensive list of security measures to be implemented when developing a Ruby on Rails application. It is designed to serve as a quick reference and minimize vulnerabilities caused by developer forgetfulness. It does not replace developer training on secure coding principles and how they can be applied.

Describing how each security vulnerability works is outside the scope of this document. Links to external resources containing further information are provided in the corresponding sections of the checklist. Please apply only the suggestions you thoroughly understand.

Please keep in mind that security is a moving target. New vulnerabilities and attack vectors are discovered every day. We suggest you try to keep up to date, for instance, by subscribing to security mailing lists related to the software and libraries you are using.

This checklist is meant to be a community-driven resource. Your contributions are welcome!

Disclaimer: This document does not cover all possible security vulnerabilities. The authors do not take any legal responsibility for the accuracy or completeness of the information herein.

Supported Rails Versions

This document focuses on Rails 4 and 5. Vulnerabilities that were present in earlier versions and fixed in Rails 4 are not included.

The Checklist
Details and Code Samples
Authors
Contributing
TODO
References and Further Reading
License

Table of contents generated by DocToc.

The Checklist

Injection

Injection attacks are #1 at the OWASP Top10.

Don’t use standard Ruby interpolation (#{foo}) to insert user inputted strings into ActiveRecord or raw SQL queries. Use the ? character, named bind variables or the ActiveRecord::Sanitization methods to sanitize user input used in DB queries. Mitigates SQL injection attacks.
Don't pass user inputted strings to methods capable of evaluating code or running O.S. commands such as eval, system, syscall, %x(), open, popen<n>, File.read, File.write, and exec. Using regular expressions is a good way to sanitize it (code sample). Mitigates command injection attacks.

Resources:

Authentication

Broken Authentication and Session Management are #2 at the OWASP Top 10.

Sessions & Cookies

Broken Authentication and Session Management are #2 at the OWASP Top 10.

Don't store data such as money/point balances or user privileges in a cookie or a CookieStore Session. Store it in the database instead. Mitigates replay attacks.
Consider always using encrypted cookies. This is the default behavior in
Rails 4+ when secret_key_base is set. Strengthens cookie encryption and mitigates multiple attacks involving cookie tampering.
Unless your JavaScript frontend needs to read cookies generated by the Rails server, set all cookies as httponly. Search the project for cookie accessors and add httponly: true. Example: cookies[:login] = {value: 'user', httponly: true}. Restricts cookie access to the Rails server. Mitigates attackers from using the victim's browser JavaScript to steal cookies after a successful XSS attack.

Resources:

Ruby on Rails Security Guide - Sessions

Cross-Site Scripting (XSS)

XSS is #3 at the OWASP Top 10.

Handling User Input

Always validate user input that may eventually be displayed to other users. Attempting to blacklist characters, strings or sanitize input tends to be ineffective (see examples of how to bypass such blacklists). A whitelisting approach is usually safer. Mitigates multiple XSS attacks.
Consider using the loofah-activerecord gem to scrub your model attribute values. Mitigates multiple XSS attacks.
If you must create links from user inputted URLs, be sure to validate them. In particular, it should be possible to limit URL schemes to http/https in nearly all cases. The URL passed to link_to (the second argument) will be HTML escaped. However, link_to allows any scheme for the URL. If using regex, ensure that the string begins with the expected protocol(s), as in \Ahttps?. Mitigates XSS attacks such as entering javascript:dangerous_stuff()//http://www.some-legit-url.com as a website URL or a dangerous data: payload that is displayed to other users (e.g., in a user profile page).
When using regex for input validation, use \A and \z to match string beginning and end. Do not use ^ and $ as anchors. Mitigates XSS attacks that involve slipping JS code after line breaks, such as me@example.com\n<script>dangerous_stuff();</script>.
Do not trust validations implemented at the client (frontend) as most implementations can be bypassed. Always (re)validate at the server.

Output Escaping & Sanitization

Escape all HTML output. Rails does that by default, but calling html_safe or raw at the view suppresses escaping. Look for calls to these methods in the entire project, check if you are generating HTML from user-inputted strings and if those strings are effectively validated. Note that there are dozens of ways to evade validation. If possible, avoid calling html_safe and raw altogether. Most templating libraries also provide a way of skipping escaping. ERB uses the double ==: <%== params[:query] %>. For custom scrubbing, see ActionView::Helpers::SanitizeHelper Mitigates XSS attacks.
Always enclose attribute values with double quotes. Even without html_safe, it is possible to introduce cross-site scripting into templates with unquoted attributes. In the following code <p class=<%= params[:style] %>...</p>, an attacker can insert a space into the style parameter and suddenly the payload is outside the attribute value and they can insert their own payload. And when a victim mouses over the paragraph, the XSS payload will fire. Mitigates XSS attacks.
Rendering JSON inside of HTML templates is tricky. You can't just HTML escape JSON, especially when inserting it into a script context, because double-quotes will be escaped and break the code. But it isn't safe to not escape it, because browsers will treat a </script> tag as HTML no matter where it is. The Rails documentation recommends always using json_escape just in case to_json is overridden or the value is not valid JSON. Mitigates XSS attacks.
Be careful when using render inline: .... The value passed in will be treated like an ERB template by default. Take a look at this code: render inline: "Thanks #{@user.name}!". Assuming users can set their own name, an attacker might set their name to <%= rm -rf / %> which will execute rm -rf / on the server! This is called Server Side Template Injection and it allows arbitrary code execution (RCE) on the server. If you must use an inline template treat all input the same as you would in a regular ERB template: render inline: "Thanks <%= @user.name %>". Mitigates XSS attacks.
Avoid sending user inputted strings in e-mails to other users. Attackers may enter a malicious URL in a free text field that is not intended to contain URLs and does not provide URL validation. Most e-mail clients display URLs as links. Mitigates XSS, phishing, malware infection and other attacks.

XSS protection in HAML templates

Be careful when using != in Haml and it should be made sure that no user data is rendered unescaped. The != notation in Haml works the way <%= raw(…) %> works in ERB. See (example code).

Resources:

Insecure Direct Object Reference

An IDOR issue arises when the user is supposed to have access to url "/get/post/6", for example, but not "/get/post/9" but the system does not properly check those permissions. And if we change "6" in the URL, what happens? We can see the data of all users. This may be due to the fact that the data was generated as follows: @user = User.find_by(id: params[:user_id]) – which is basically getting the ID from the GET parameter in the URL. Instead a more secure way of doing this is setting the @user parameter based on the "current_user" session variable like this: @user = current_user.

Resources:

Rails Vulnerabilities and Where To Find Them – Part 1

HTTP & TLS

Force HTTPS over TLS (formerly known as SSL). Set config.force_ssl = true in config/environments/production.rb. May also be done in a TLS termination point such as a load balancer, Nginx or Passenger Standalone. Mitigates man-in-the-middle and other attacks.
Use the SSL Server Test tool from Qualys SSL Lab to check the grade of your TLS certificate. Be sure to use the strongest (yet widely compatible) protocols and cipher suites, preferably with Ephemeral Diffie-Hellman support. The
Mozilla SSL Configuration Generator can give you some suggestions. Mitigates multiple SSL/TLS-related attacks such as BEAST and POODLE.
Consider rate-limiting incoming HTTP requests, as implemented by the rack-attack and rack-throttle gems. See sample code. Mitigates web scraping, HTTP floods, and other attacks.

Security-related headers

Consider using the Secure Headers gem. Mitigates several attacks.
Consider obfuscating the web server banner string. In other words, hide your web server name and version. Mitigates HTTP fingerprinting, making it harder for attackers to determine which exploits may work on your web server.

Memcached Security

Use a firewall. Memcached needs to be accessible from your other servers but there's no reason to expose it to the internet. In short, only your other production servers have access to your production memcached servers. This alone would prevent your server from being used in an attack. Memcached out of the box doesn't use authentication so anyone who can connect to your server will be able to read your data.
Listen on a private interface. If you're running one server for your Rails application and memcached, you should listen on 127.0.0.1. For availability reasons, you shouldn't have 1 server in production anyway. For staging and test environments, follow this rule. For production setups where you have multiple Rails servers that need to connect to memcached, use the private IP of the server. This is something like 192.168.0.1, 172.16.0.1, or 10.0.0.1. When you start memcached, use --listen 127.0.0.1 or --listen 192.168.0.1.
Disable UDP. It is enabled by default. To disable UDP, use -U 0 when starting memcached.

Resources:

Memcached Security aka Don't Attack GitHub

Authorization (Pundit)

Implement authorization at the back end. Hiding links/controls in the UI is not enough to protect resources against unauthorized access. Mitigates forced browsing attacks.
Ensure all controllers/actions which require authorization call the authorize or policy_scope method (sample code). Mitigates forced browsing attacks due to developers forgetting to require authorization in some controller actions.
When using DB records associated to users to populate select boxes, radio buttons or checkboxes, instead of querying by association (user.posts), consider using policy_scope. See additional details and sample code. Improves readability and maintainability of authorization policies.

Resources:

Files

File Uploads

File Downloads

Do not allow downloading of user-submitted filenames and paths. If not possible, use a whitelist of permitted filenames and paths. Mitigates the exploitation of directory traversal vulnerabilities to download sensitive files.

Resources:

Ruby on Rails Security Guide - File Uploads and Downloads

Cross-Site Request Forgery (CSRF)

Enforce CSRF protection by setting protect_from_forgery with: :exception in all controllers used by web views or in ApplicationController.
Use HTTP verbs in a RESTful way. Do not use GET requests to alter the state of resources. Mitigates CSRF attacks.
Up to Rails 4, there was a single CSRF token for all forms, actions, and methods. Rails 5 implements per-form CSRF tokens, which are only valid for a single form and action/method. Enable it by setting config.action_controller.per_form_csrf_tokens = true.

Resources:

Cross Origin Resource Sharing (CORS)

Occasionally the need to share some resources across many domains appears. For example, you want to upload a file using AJAX request and send it to the other app. The receiving side should specify a whitelist of domains that are allowed to make those requests. There are few HTTP headers that control that.

You can use rack-cors gem and in config/application.rb specify your configuration (code sample).

Resources:

Security issues solutions in RoR

Sensitive Data Exposure

If possible, avoid storing sensitive data such as credit cards, tax IDs and third-party authentication credentials in your application. If not possible, ensure that all sensitive data is encrypted at rest (in the DB) and in transit (use HTTPS over TLS). Mitigate theft/leakage of sensitive data.
Do not log sensitive data such as passwords and credit card numbers. You may include parameters that hold sensitive data in config.filter_parameters at initializers/filter_parameter_logging.rb. For added security, consider converting filter_parameters into a whitelist. See sample code. Prevents plain-text storage of sensitive data in log files.
HTML comments are viewable to clients and should not contain details that can be useful to attackers. Consider using server-side comments such as <%# This comment syntax with ERB %> instead of HTML comments. Avoids exposure of implementation details.
Avoid exposing numerical/sequential record IDs in URLs, form HTML source and APIs. Consider using slugs (A.K.A. friendly IDs, vanity URLs) to identify records instead of numerical IDs, as implemented by the friendly_id gem. Additional benefits include SEO and better-looking URLs. Mitigates forced browsing attacks and exposure of metrics about your business, such as the number of registered users, number of products on stock, or number of receipts/purchases.
If using slugs instead of numerical IDs for URLs, consider returning a 404 Not Found status code instead of 403 Forbidden for authorization errors. Prevents leakage of attribute values used to generate the slugs. For instance, visiting www.myapp.com/users/john-doe and getting a 403 return status indicates the application has a user named John Doe.*
Do not set config.consider_all_requests_local = true in the production environment. If you need to set config.consider_all_requests_local = true to use the better_errors gem, do it on config/environments/development.rb. Prevents leakage of exceptions and other information that should only be accessible to developers.
Don't install development/test-related gems such as better_errors and web-console in the production environment. Place them within a group :development, :test do block in the Gemfile. Prevents leakage of exceptions and even REPL access if using better_errors + web-console.

Credentials

The encryption key, located on config/master.key is created when you run rails new. It's also added to .gitignore so it doesn't get committed to your repository. Mitigates credential leaks/theft.
Don't edit the config/credentials.yml.enc file directly. To add credentials, run bin/rails credentials:edit. Use a flat format which means you don't have to put development or production anymore. Mitigates credential leaks/theft.
If you want to generate a new secret key base run, bin/rails secret and add that to your credentials by running bin/rails credentials:edit.
Upload master.key securely. You can scp or sftp the file. Upload the key to a shared directory. Shared here means shared between releases, not a shared filesystem. On each deploy, you symlink config/master.key to /path/to/shared/config/master.key.
If you need to give a developer a copy of the key, never send it via email (unless you're using encrypted emails which most of us don't!) You can use a password manager because they use encryption.
Put the key on the RAILS_MASTER_KEY environment variable. In some cases where you can't upload a file, this is the only option. Even though this is convenient, make sure you know the risks of using environment variables. The risks can be mitigated, but if you can upload master.key then use that option.

Resources:

Rails Encrypted Credentials on Rails 5.2

Routing, Template Selection, and Redirection

Don't perform URL redirection based on user inputted strings. In other words, don't pass user input to redirect_to. If you have no choice, create a whitelist of acceptable redirect URLs or limit to only redirecting to paths within your domain (example code). Mitigates redirection to phishing and malware sites. Prevent attackers from providing URLs such as http://www.my-legit-rails-app.com/redirect?to=www.dangeroussite.com to victims.
Do not use a user inputted string to determine the name of the template or view to be rendered. Prevents attackers from rendering arbitrary views such as admin-only pages.
Avoid "catch-all" routes such as match ':controller(/:action(/:id(.:format)))' and make non-action controller methods private. Mitigates unintended access to controller methods.

Resources:

OWASP Ruby on Rails Cheatsheet - Redirects and Forwards (URL validation)

Third-party Software

Apply the latest security patches in the OS frequently. Pay special attention to internet-facing services such as application servers (Passenger, Puma, Unicorn), web servers (Nginx, Apache, Passenger Standalone) and SSH servers.
Update Ruby frequently.
Watch out for security vulnerabilities in your gems. Run bundler-audit frequently or use a service like Snyk, Gemnasium (both free for open-source development) or Appcanary.

Security Tools

Run Brakeman before each deploy. If using an automated code review tool like Code Climate, enable the Brakeman engine.
Consider using a continuous security service such as Detectify.
Consider using a Web Application Firewall (WAF) such as NAXSI for Nginx, ModSecurity for Apache and Nginx. Mitigates XSS, SQL Injection, DoS, and many other attacks.

Testing

Include security tests in your test suite. Look at OWASP's RailsGoat application for examples of security-related Capybara specs. Raises additional security awareness and mitigates security-related regressions.
Create security tests in pairs: one for the access denied scenario and another for the access granted scenario.
When using TDD, consider implementing authentication in the early stages of development, as it tends to break multiple preexisting tests.

Others

Use strong parameters in the controllers. This is the default behavior as of Rails 4+. Mitigates mass assignment attacks such as overwriting the role attribute of the User model for privilege escalation purposes.
Implement Captcha or Negative Captcha on publicly exposed forms. reCAPTCHA is a great option, and there is a gem that facilitates Rails integration. Other options are the rucaptcha and negative-captcha gems. Mitigates automated SPAM (spambots).

Details and Code Samples

Command Injection example

# User input
params[:shop][:items_ids] # Maybe you expect this to be an array inside a string.
                          # But it can contain something very dangerous like:
                          # "Kernel.exec('Whatever OS command you want')"

# Vulnerable code
evil_string = params[:shop][:items_ids]
eval(evil_string)

If you see a call to eval you must be very sure that you are properly sanitizing it. Using regular expressions is a good way to accomplish that.

# Secure code
evil_string = params[:shop][:items_ids]
secure_string = /\[\d*,?\d*,?\d*\]/.match(evil_string).to_s

eval(secure_string)

Password validation regex

We may implement password strength validation in Devise by adding the following code to the User model.

validate :password_strength

private

def password_strength
  minimum_length = 8
  # Regex matches at least one lower case letter, one uppercase, and one digit
  complexity_regex = /\A(?=.*[a-z])(?=.*[A-Z])(?=.*[0-9])/
  # When a user is updated but not its password, the password param is nil
  if password.present? &&
    (password.length < minimum_length || !password.match(complexity_regex))
    errors.add :password, 'must be 8 or more characters long, including 
                           at least one lowercase letter, one uppercase
                           letter, and one digit.'
  end
end

Pundit: ensure all actions are authorized

Add the following to app/controllers/application_controller.rb

after_action :verify_authorized, except: :index, unless: :devise_controller?
after_action :verify_policy_scoped, only: :index, unless: :devise_controller?

Add the following to controllers that do not require authorization. You may create a concern for DRY purposes.

after_action_skip :verify_authorized
after_action_skip :verify_policy_scoped

Pundit: only display appropriate records in select boxes

Think of a blog-like news site where users with editor role have access to specific news categories, and admin users have access to all categories. The User and the Category models have an HMT relationship. When creating a blog post, there is a select box for choosing a category. We want editors only to see their associated categories in the select box, but admins must see all categories. We could populate that select box with user.categories. However, we would have to associate all admin users with all categories (and update these associations every time a new category is created). A better approach is to use Pundit Scopes to determine which categories are visible to each user role and use the policy_scope method when populating the select box.

# app/views/posts/_form.html.erb
f.collection_select :category_id, policy_scope(Category), :id, :name

Convert filter_parameters into a whitelist

Developers may forget to add one or more parameters that contain sensitive data to filter_parameters. Whitelists are usually safer than blacklists as they do not generate security vulnerabilities in case of developer forgetfulness. The following code converts filter_parameters into a whitelist.

# config/initializers/filter_parameter_logging.rb
if Rails.env.production?
  # Parameters whose values are allowed to appear in the production logs:
  WHITELISTED_KEYS = %w(foo bar baz)
  
  # (^|_)ids? matches the following parameter names: id, *_id, *_ids
  WHITELISTED_KEYS_MATCHER = /((^|_)ids?|#{WHITELISTED_KEYS.join('|')})/.freeze
  SANITIZED_VALUE = '[FILTERED]'.freeze
  
  Rails.application.config.filter_parameters << lambda do |key, value|
    unless key.match(WHITELISTED_KEYS_MATCHER)
      value.replace(SANITIZED_VALUE)
    end
  end
else
  # Keep the default blacklist approach in the development environment
  Rails.application.config.filter_parameters += [:password]
end

rack-cors configuration

module Sample
  class Application < Rails::Application
    config.middleware.use Rack::Cors do
      allow do
        origins 'someserver.example.com'
        resource %r{/users/\d+.json},
          headers: ['Origin', 'Accept', 'Content-Type'],
          methods: [:post, :get]
      end
    end
  end
end

Throttling Requests

On some pages like the login page, you'll want to throttle your users to a few requests per minute. This prevents bots from trying thousands of passwords quickly.

Rack Attack is a Rack middleware that provides throttling among other features.

Rack::Attack.throttle('logins/email', :limit => 6, :period => 60.seconds) do |req|
  req.params['email'] if req.path == '/login' && req.post?
end

HAML: XSS protection

By default,

="<em>emphasized<em>"
!= "<em>emphasized<em>"

compiles to:

&lt;em&gt;emphasized&lt;/em&gt;
<em>emphasized<em>

Authors

Bruno Facca - LinkedIn - Email: bruno at facca dot info

Contributing

Contributions are welcome. If you would like to correct an error or add new items to the checklist, feel free to create an issue followed by a PR. See the TODO section for contribution suggestions.

If you are interested in contributing regularly, drop me a line at the above e-mail to become a collaborator.

TODO

Add sample tests (RSpec and/or Minitest) to detect the presence of vulnerabilities. See OWASP's RailsGoat security-related Capybara specs for inspiration.
Compare upload gems regarding their implementation of the File Uploads items of this checklist (build a table).
Compare authentication gems regarding their implementation of the Authentication items of this checklist (build a table).

References and Further Reading

License

Released under the MIT License.

christophwong/zen-rails-security-checklist

Zen Rails Security Checklist

Summary

Supported Rails Versions

Table of Contents

The Checklist

Injection

Authentication

Sessions & Cookies

Cross-Site Scripting (XSS)

Handling User Input

Output Escaping & Sanitization

XSS protection in HAML templates

Insecure Direct Object Reference

HTTP & TLS

Security-related headers

Memcached Security

Authorization (Pundit)

Files

File Uploads

File Downloads

Cross-Site Request Forgery (CSRF)

Cross Origin Resource Sharing (CORS)

Sensitive Data Exposure

Credentials

Routing, Template Selection, and Redirection

Third-party Software

Security Tools

Testing

Others

Details and Code Samples

Command Injection example

Password validation regex

Pundit: ensure all actions are authorized

Pundit: only display appropriate records in select boxes

Convert filter_parameters into a whitelist

rack-cors configuration

Throttling Requests

HAML: XSS protection

Authors

Contributing

TODO

References and Further Reading

License