Share
Embed
Transcript
49 slides
-
Overview • Authorization Attacks • Injection Attacks • SQL Injection
• Cross-Site Scripting • Cross-Site Request Forgery Attacks • Mass Assignment Attacks -
Examples • The examples in this presentation are based on
a simple social network application • Users sign up for accounts and create posts • Assume User and Post models • A user has many posts • A post belongs to a user -
Definitions • Authentication identifies a user • Authorization specifies what
a logged-in user can access within your application -
Authorization Attack • This attack is sometimes called an Insecure
Direct Object Reference • In other words, a user might change the URL from posts/1 to posts/2 and try to edit another user’s post -
Example # This method lets any logged-in # user update
any post. def update @post = Post.find(params[:id]) # ... end -
Example # This method ensures that a user # can
only update their own post def update @post = current_user.posts.find(params[:id]) # ... end -
Injection Attacks • An injection attack occurs when user input
is executed as part of the application • Injection attacks are extremely common, especially in older applications • The first rule of avoiding injection attacks is never trust input from the user -
SQL Injection # Never trust user input class User def
self.authenticate(username, password) where("username = '#{username}'" + " AND password = '#{password}'").first end end -
So Far, So Good User.authenticate("tony", "secret") SELECT * FROM users
WHERE (username = 'tony' AND password = 'secret') LIMIT 1 => #<User id: 1, username: ...> -
So Far, So Good User.authenticate("tony", "wrong") SELECT * FROM users
WHERE (username = 'tony' AND password = 'wrong') LIMIT 1 => nil -
Magic Words • Attackers know a magic phrase for exploiting
SQL Injection vulnerabilities. • What if we enter this for the password? ' OR '1' = '1 -
Uh Oh User.authenticate("tony", "' OR '1' = '1") SELECT *
FROM users WHERE (username = 'tony' AND password = '' OR '1' = '1') LIMIT 1 => #<User id: 1, username: ...> -
What Happened? • My password isn’t an empty string, but
the addition of OR '1' = '1' made the expression true • The attacker is now logged in as me -
Solution # Switch to hash conditions class User def self.authenticate(username,
password) where(username: username, password: password).first end end -
Solution • Rails automatically SQL escapes strings in hash conditions
• If you see string interpolation inside a where method, assume it is dangerous -
Cross-Site Scripting • A vulnerability that allows attackers to inject
client- side JavaScript into web pages viewed by others • When another user views the page, the JavaScript is executed as if it were part of your application -
Safe By Default • Thankfully, Rails HTML escapes all output
by default • Symbols such as < are replaced with character entities such as < • But what if your application needs to show HTML output? -
Example # Create a new post with # JavaScript in
the body Post.create( title: "XSS Test", body: "<script>alert('XSS');</script>" ) -
Example # This view code... <%= @post.body %> # Becomes
this HTML... <script>alert('XSS');</script> # This will not execute -
Safe By Default • But what if your application needs
to show HTML output? • Users want to use HTML to format their posts -
Example # This view code... <%= raw @post.body %> #
Becomes this HTML... <script>alert('XSS');</script> # This WILL execute -
Remove Unsafe Tags • Rails includes the sanitize method to
automatically remove unsafe HTML tags • The sanitize method uses a whitelist of allowed tags to clean up potentially unsafe strings • In Rails 4.2, sanitize is based on the Loofah gem -
Example # Use helper.sanitize in a console helper.sanitize "<p>Hello</p>" =>
"<p>Hello</p>" # pass a list of allowed tags tags = ['em', 'strong'] helper.sanitize "<p>Hello</p>", tags: tags => "Hello" -
Example # This view code... <%= sanitize @post.body %> #
Becomes an empty string... # Obviously, this will not execute -
CSRF Attacks • Occurs when one of your application’s users
visits another site that has been modified by an attacker to target your site • The attacker must trick your application’s users into visiting the malicious page and activating the link • The malicious page uses your application’s trust in this user to accept requests -
Example • Imagine you are building an online payment application
• Your application includes a transfer action that accepts amount and to parameters that specify how much money to transfer to another user • An attacker studies the requests generated by your site and attempts to replicate those requests on his own site -
Example <!-- an image tag on the attacker’s site causes
a GET request --> <img src="http://yoursite.com/transfer? amount=100&to=attacker"> -
Example • If one of your users visits this site,
and your site is vulnerable to CSRF attacks, $100 is transferred from their account to the attackers account • This could also be done with a script on the attacker’s site that issues requests -
CSRF Prevention • Never use a GET request to change
state • Use a POST request to update data • Include a user-specific token with all requests that change state in your application • Ignore requests that do not include this token -
CSRF Prevention • Rails includes helper methods for generating a
user-specific token automatically • Rails calls this an authenticity_token • The token is stored in a meta tag on each page • The token is also included in a hidden field on every form -
CSRF Prevention # Your application layout includes # a per-user
token by default... <%= csrf_meta_tags %> # That renders this HTML <meta content="authenticity_token" name="csrf-param" /> <meta content="KA1Q/JoVfI+aV6/L4..." name="csrf-token" /> -
CSRF Prevention # Your application controller raises # an exception
if the token is not # included with a POST request... protect_from_forgery with: :exception -
Mass Assignment • Model methods such as update_attributes and new
accept a hash of attributes and update or set all attributes by default • Older Rails versions used the attr_accessible model method to define attributes that could be mass assigned • Rails 4 uses the Strong Parameters gem to move this responsibility to the controller -
Example • An attacker can modify your application’s forms or
use a tool such as cURL to send parameters you aren’t expecting • For example, the attacker might try sending admin=true to your user signup form -
Example # The params variable in your # controller now
looks like this: params = { user: { name: "Tony", admin: true } } -
Example # In older Rails versions this assigns # all
attributes unless attr_accessible # is setup on the User model @user = User.new(params[:user]) # Assuming your User model has an admin # flag, the new user is now an admin -
Example # In Rails 4 this raises an exception #
because it is a mass assignment # without a call to permit @user = User.new(params[:user]) # The new user is not created -
Example # The preferred pattern is to call # a
private user_params method that # uses the permit and require methods # to setup allowed params @user = User.new(user_params) -
Example private def user_params params.require(:user).permit(:name, :email) end # this method
returns this hash: # { name: "Tony" } # the admin flag is silently removed -
Summary • As your application gains popularity, the risk of
attack rises • Rails provides the tools you need to protect your application and your users from attacks • Stay up-to-date on security issues -
Resources • The examples in this presentation are based on
Chapter 11 of my book Rails Crash Course • The security vulnerabilities were taken from the OWASP Top 10 List • The Open Web App Security Project • www.owasp.org