End-To-End Test Your Web Security

Gleb Bahmutov

Sr Director of Engineering, Mercari US

survival is possible* but we need to act now

  • change your life
  • join an organization

Speaker: Gleb Bahmutov PhD

C / C++ / C# / Java / CoffeeScript / JavaScript / Node / Angular / Vue / Cycle.js / functional programming / testing

Gleb Bahmutov

Sr Director of Engineering

glebbahmutov.com/blog testing posts

Lots Of Testing

Enter your username / password / credit card

You can have your

website hacked but only once.

Security, security, security

Open Worldwide Application Security Project (OWASP)

https://owasp.org/www-project-top-ten/

XSS / code injection attacks

Code injection attacks

=

confusion between data and code

<html>
  <head>
    <style>...</style>
    <script src="https://code.jquery.com/3.5.0/jquery.min.js"></script>
    <script>
      ...
    </script>
  </head>
  <body>
    <h1>Chat</h1>
    <ul>
      <li>Chat message 1</li>
      <li>Chat message 2</li>
    </ul>
  </body>
</html>
<html>
  <head>
    <style>...</style>
    <script src="https://code.jquery.com/3.5.0/jquery.min.js"></script>
    <script>
      ...
    </script>
  </head>
  <body>
    <h1>Chat</h1>
    <ul>
      <li>Chat message 1</li>
      <li>Chat message 2</li>
    </ul>
  </body>
</html>

Data

<html>
  <head>
    <style>...</style>
    <script src="https://code.jquery.com/3.5.0/jquery.min.js"></script>
    <script>
      ...
    </script>
  </head>
  <body>
    <h1>Chat</h1>
    <ul>
      <li>Chat message 1</li>
      <li>Chat message 2</li>
    </ul>
  </body>
</html>

Code

Name:

John Doe

<span>{{ name }}<span>

<span>John Doe<span>

What about user input?

Name:

<script>alert('hi')</script>

<span>{{ name }}<span>

<span><script>alert('hi')</script><span>

innerText or innerHTML?

What about user input?

do you use eval(user input) anywhere???

Name:

<script>alert('hi')</script>

<span>{{ name }}<span>

<span><script>alert('hi')</script><span>

innerText or innerHTML?

What about user input?

do you use eval(user input) anywhere???

User input (data) is treated as code

innerText or innerHTML?

Dangerous, but has styling, elements, etc

Really dangerous if one user's content

can be viewed by another user

Will a script embedded in the user-entered content be executed (for another user)?

NEVER EVER EVER?

My Site

My content

 

 

 

 

 

 

 

 

User comments

Ads, etc

 

 

 

 

 

 

 

<script>stealStuff()</script>

User input can come from:

  • input elements
  • database
  • URL
  • ...

Cross-Site Scripting attack (XSS)

Let's Test It!

To test:

  • happy user path
  • unsuccessful injection
  • successful injection
  • injection prevention and reporting
// app.js
send$.addEventListener('click', () => {
  const message = message$.value
  const li = document.createElement('li')
  li.innerText = message
  messages$.appendChild(li)
  message$.value = ''
  send$.setAttribute('disabled', 'disabled')
})

// innerText (safe)

// cypress/e2e/spec.cy.js
it('adds a new message', () => {
  cy.visit('/')
  cy.contains('button', 'Send').should('be.disabled')
  cy.get('#message').type('Hello')
  cy.contains('button', 'Send').click()
  cy.get('#messages li')
    .should('have.length', 1).and('have.text', 'Hello')
  cy.get('#message').should('have.value', '')
  cy.contains('button', 'Send').should('be.disabled')
})

Cypress end-to-end test

https://docs.cypress.io/

// cypress/e2e/spec.cy.js
it('adds a new message', () => {
  cy.visit('/')
  cy.contains('button', 'Send').should('be.disabled')
  cy.get('#message').type('Hello')
  cy.contains('button', 'Send').click()
  cy.get('#messages li')
    .should('have.length', 1).and('have.text', 'Hello')
  cy.get('#message').should('have.value', '')
  cy.contains('button', 'Send').should('be.disabled')
})

Cypress end-to-end test

https://docs.cypress.io/

Cypress tests commands slowed by 1 second each using cypress-slow-down plugin

Estimate: 3 points

Priority: P0

// app.js
send$.addEventListener('click', () => {
  const message = message$.value
  const li = document.createElement('li')
  li.innerHTML = message
  messages$.appendChild(li)
  message$.value = ''
  send$.setAttribute('disabled', 'disabled')
})

// innerHTML (for now)

// cypress/e2e/spec.cy.js
it('adds a bold message', () => {
  cy.visit('/')
  cy.get('#message').type('<b>Important</b>')
  cy.contains('button', 'Send').click()
  cy.get('#messages li b')
    .should('have.length', 1)
    .and('have.text', 'Important')
})
// cypress/e2e/spec.cy.js
it('adds a bold message', () => {
  cy.visit('/')
  cy.get('#message').type('<b>Important</b>')
  cy.contains('button', 'Send').click()
  cy.get('#messages li b')
    .should('have.length', 1)
    .and('have.text', 'Important')
})

Testing <script>... inputs

// cypress/e2e/spec.cy.js
it('injecting <script> tag does not work', () => {
  cy.on('window:load', (win) => cy.stub(win.console, 'log').as('log'))
  cy.visit('/')
  cy.get('#message').type('Hello<script>console.log(`hacked`)</script>')
  cy.contains('button', 'Send').click()
  cy.contains('#messages li', 'Hello')
  cy.get('@log').should('not.have.been.called')
})
// cypress/e2e/spec.cy.js
it('injecting <script> tag does not work', () => {
  cy.on('window:load', (win) => cy.stub(win.console, 'log').as('log'))
  cy.visit('/')
  cy.get('#message').type('Hello<script>console.log(`hacked`)</script>')
  cy.contains('button', 'Send').click()
  cy.contains('#messages li', 'Hello')
  cy.get('@log').should('not.have.been.called')
})

Testing <script>... inputs

Are we 100% safe?!

// cypress/e2e/spec.cy.js
it('injects XSS via img onerror attribute', () => {
  cy.on('window:load', (win) => 
        cy.stub(win.console, 'log').as('log'))
  cy.visit('/')
  cy.get('#message')
    .type('Hello<img src="" onerror="console.log(`hacked`)" />')
  cy.contains('button', 'Send').click()
  cy.contains('#messages li', 'Hello')
  cy.get('@log').should('have.been.calledWith', 'hacked')
})
// cypress/e2e/spec.cy.js
it('injects XSS via img onerror attribute', () => {
  cy.on('window:load', (win) => 
        cy.stub(win.console, 'log').as('log'))
  cy.visit('/')
  cy.get('#message')
    .type('Hello<img src="" onerror="console.log(`hacked`)" />')
  cy.contains('button', 'Send').click()
  cy.contains('#messages li', 'Hello')
  cy.get('@log').should('have.been.calledWith', 'hacked')
})

Are we 100% safe?!

Are we 100% safe?!

No, we are not

Strip all but <b> and <i> tags... ?

// app.js
send$.addEventListener('click', () => {
  const message = message$.value
  const li = document.createElement('li')
  li.innerHTML = removeBadTags(message)
  messages$.appendChild(li)
  message$.value = ''
  send$.setAttribute('disabled', 'disabled')
})

Strip all but <b> and <i> tags... ?

it('injects XSS via b onclick attribute', () => {
  cy.on('window:load', (win) => 
        cy.stub(win.console, 'log').as('log'))
  cy.visit('/')
  cy.get('#message')
    .type('<b onclick="console.log(`hacked`)">Hello</b>')
  cy.contains('button', 'Send').click()
  cy.contains('#messages li b', 'Hello').click()
  cy.get('@log').should('have.been.calledWith', 'hacked')
})

Strip all but <b> and <i> tags... ?

it('injects XSS via b onclick attribute', () => {
  cy.on('window:load', (win) => 
        cy.stub(win.console, 'log').as('log'))
  cy.visit('/')
  cy.get('#message')
    .type('<b onclick="console.log(`hacked`)">Hello</b>')
  cy.contains('button', 'Send').click()
  cy.contains('#messages li b', 'Hello').click()
  cy.get('@log').should('have.been.calledWith', 'hacked')
})

Will never happen to me

Angular 1 edition (dated)

Examples where we can embed script code into {{ }} template

{{constructor.constructor('alert(1)')()}}

fixed in Angular 1.2

Will never happen to me

{{ (_=''.sub).call.call({}[$='constructor'].getOwnPropertyDescriptor ( _.__proto__,$).value,0,'alert(1)')() }}

Will never happen to me

{{ objectPrototype = ({})[['__proto__']]; objectPrototype[['__defineSetter__']]('$parent', $root.$$postDigest); $root.$$listenerCount[['constructor']] = 0; $root.$$listeners = [].map; $root.$$listeners.indexOf = [].map.bind; functionPrototype = [].map[['__proto__']]; functionToString = functionPrototype.toString; functionPrototype.push = ({}).valueOf; functionPrototype.indexOf = [].map.bind; foo = $root.$on('constructor', null); functionPrototype.toString = $root.$new; foo(); }} {{ functionPrototype.toString = functionToString; functionPrototype.indexOf = null; functionPrototype.push = null; $root.$$listeners = {}; baz ? 0 : $root.$$postDigestQueue[0]('alert(location)')(); baz = true;'' }} 

Will never happen to me

{{ 'this is how you write a number properly. also, numbers are basically arrays.'; 0[['__proto__']].toString = [][['__proto__']].pop; 0[['__proto__']][0] = 'alert("TROLOLOLn"+document.location)'; 0[['__proto__']].length = 1; 'did you know that angularjs eval parses, then re-stringifies numbers? :)'; $root.$eval("x=0", $root); }}

Lesson: you cannot sanitize your way out of <script> tags

At least: do not write the sanitizer yourself, use https://github.com/cure53/DOMPurify

Estimate: 1 point

Priority: P2

Solution: disable inline JavaScript

<body>
  <script>function willNotRun() {...}</script>
  <script src="assets/js/script.js"></script>
</body>

allowed from the right domain

never* allowed

* or with specific exceptions

Attacker (probably) cannot change HTML / JavaScript on our server or replace a CDN library*

* tip: use exact library URLs and integrity checksums

Content-Security-Policy (CSP)

<meta http-equiv="Content-Security-Policy" 
      content="script-src https://code.jquery.com 'self';">

Wide support for CSP (browsers, web apps)

or even better

Content-Security-Policy: default-src 'self' example.com *.example.com

Content-Security-Policy (CSP)

Content-Security-Policy: default-src 'self'; img-src *; media-src example.org example.net; script-src userscripts.example.com

Specify trusted sources for scripts, fonts, styles, images, media, frames, etc

Set CSP Header

// server JS
const express = require('express')
const helmet = require('helmet')
const app = express()
app.use(
  helmet({
    contentSecurityPolicy: {
      directives: {
        defaultSrc: ["'self'"],
        // ughh, have to allow unsafe inline styles to be added by Cypress
        // https://github.com/cypress-io/cypress/issues/21374
        styleSrc: ["'self'", "'unsafe-inline'"],
        reportUri: ['/security-attacks'],
      },
    },
  }),
)
// cypress.config.js
// https://on.cypress.io/experiments
// https://github.com/cypress-io/cypress/issues/1030
experimentalCspAllowList: ['default-src', 'script-src'],
  
// cypress/e2e/spec.cy.js
it('stops XSS', () => {
  cy.on('window:load', (win) => 
        cy.stub(win.console, 'log').as('log'))
  cy.visit('/')
  cy.get('#message')
    .type('Hello<img src="" onerror="console.log(`hacked`)" />')
  cy.contains('button', 'Send').click()
  cy.contains('#messages li', 'Hello')
  cy.get('@log').should('not.be.called')
})
// cypress.config.js
// https://on.cypress.io/experiments
// https://github.com/cypress-io/cypress/issues/1030
experimentalCspAllowList: ['default-src', 'script-src'],
  
// cypress/e2e/spec.cy.js
it('stops XSS', () => {
  cy.on('window:load', (win) => 
        cy.stub(win.console, 'log').as('log'))
  cy.visit('/')
  cy.get('#message')
    .type('Hello<img src="" onerror="console.log(`hacked`)" />')
  cy.contains('button', 'Send').click()
  cy.contains('#messages li', 'Hello')
  cy.get('@log').should('not.be.called')
})

CSP Violation error

CSP Violation Reports

CSP Violation Reports

CSP Violation Reports

// server JS
const express = require('express')
const helmet = require('helmet')
const app = express()
app.use(
  helmet({
    contentSecurityPolicy: {
      directives: {
        defaultSrc: ["'self'"],
        // ughh, have to allow unsafe inline styles to be added by Cypress
        // https://github.com/cypress-io/cypress/issues/21374
        styleSrc: ["'self'", "'unsafe-inline'"],
        reportUri: ['/security-attacks'],
      },
    },
  }),
)

Test CSP violation reporting

// cypress/e2e/spec.cy.js
it('stops XSS and reports CSP violations', () => {
  cy.intercept('/security-attacks', {}).as('cspAttacks')

  cy.on('window:load', (win) => cy.stub(win.console, 'log').as('log'))
  cy.visit('/')
  cy.get('#message').type('Hello<img src="" onerror="console.log(`hacked`)" />')
  cy.contains('button', 'Send').click()
  cy.contains('#messages li', 'Hello')

  cy.log('**XSS stopped and reported**')
  cy.wait('@cspAttacks').its('request.body').should('include', 'blocked')
  cy.get('@log').should('not.be.called')
})

Test CSP violation reporting

// cypress/e2e/spec.cy.js
it('stops XSS and reports CSP violations', () => {
  cy.intercept('/security-attacks', {}).as('cspAttacks')

  cy.on('window:load', (win) => cy.stub(win.console, 'log').as('log'))
  cy.visit('/')
  cy.get('#message').type('Hello<img src="" onerror="console.log(`hacked`)" />')
  cy.contains('button', 'Send').click()
  cy.contains('#messages li', 'Hello')

  cy.log('**XSS stopped and reported**')
  cy.wait('@cspAttacks').its('request.body').should('include', 'blocked')
  cy.get('@log').should('not.be.called')
})

Code injection attacks are common

Use content security policy

Write tests to check your CSP directives

Verify the violation reporting works

Learn More

💡 Trusted types experimental browser feature https://web.dev/trusted-types/

Gleb Bahmutov

gleb.dev

👏 Thank You 👏

End-To-End Test Your Web Security

End-To-End Test Your Web Security

By Gleb Bahmutov

End-To-End Test Your Web Security

Script injection attacks can load the attacker's code and run it on your website when other users browse it. You can prevent such attacks using content security policies, but how do you ensure your defense mechanisms actually work? By testing them! In this talk, I explain the content-security-policy, the security violation reporting, and how we can write a Cypress test to verify the attacks are stopped.

  • 949