Securing server-rendered applications

If you are creating a server-rendered application using AdonisJS, then you must use the @adonisjs/shield package to protect your applications from common web attacks like CSRF, XSS, Content sniffing, and so on.

The package comes pre-configured with the web starter kit. However, you can manually install and configure the package as follows.

The @adonisjs/shield package has a peer dependency on the @adonisjs/session package, so make sure to configure the session package first.

node ace add @adonisjs/shield

Installs the @adonisjs/shield package using the detected package manager.

Registers the following service provider inside the adonisrc.ts file.

{
  providers: [
    // ...other providers
    () => import('@adonisjs/shield/shield_provider'),
  ]
}

Creates the config/shield.ts file.
Registers the following middleware inside the start/kernel.ts file.
```
router.use([() => import('@adonisjs/shield/shield_middleware')])
```

CSRF protection

CSRF (Cross-Site Request Forgery) is an attack in which a malicious website tricks the users of your web app to perform form submissions without their explicit consent.

To protect against CSRF attacks, you should define a hidden input field holding the CSRF token value that only your website can generate and verify. Hence, the form submissions triggered by the malicious website will fail.

Protecting forms

Once you configure the @adonisjs/shield package, all form submissions without a CSRF token will automatically fail. Therefore, you must use the csrfField edge helper to define a hidden input field with the CSRF token.

Edge helper

<form method="POST" action="/">
  {{ csrfField() }}
  <input type="name" name="name" placeholder="Enter your name">
  <button type="submit"> Submit </button>
</form>

Output HTML


<form method="POST" action="/">
    <input type="hidden" name="_csrf" value="Q9ghWSf0-3FD9eCiu5YxvKaxLEZ6F_K4DL8o"/>
    <input type="name" name="name" placeholder="Enter your name"/>
    <button type="submit">Submit</button>
</form>

During the form submission, the shield_middleware will automatically verify the _csrf token, only allowing the form submissions with a valid CSRF token.

Handling exceptions

Shield raises an E_BAD_CSRF_TOKEN exception when the CSRF token is missing or invalid. By default, AdonisJS will capture the exception and redirect the user back to the form with an error flash message.

You can access the flash message as follows inside an edge template.

@error('E_BAD_CSRF_TOKEN')
  <p> {{ $message }} </p>
@end

<form method="POST" action="/">
  {{ csrfField() }}
  <input type="name" name="name" placeholder="Enter your name">
  <button type="submit"> Submit </button>
</form>

You can also self-handle the E_BAD_CSRF_TOKEN exception inside the global exception handler as follows.

import app from '@adonisjs/core/services/app'
import { errors } from '@adonisjs/shield'
import { HttpContext, ExceptionHandler } from '@adonisjs/core/http'

export default class HttpExceptionHandler extends ExceptionHandler {
  async handle(error: unknown, ctx: HttpContext) {
    if (error instanceof errors.E_BAD_CSRF_TOKEN) {
      return ctx.response
        .status(error.status)
        .send('Page has expired')
    }
    return super.handle(error, ctx)
  }
}

Config reference

The configuration for the CSRF guard is stored inside the config/shield.ts file.

import { defineConfig } from '@adonisjs/shield'

const shieldConfig = defineConfig({
  csrf: {
    enabled: true,
    exceptRoutes: [],
    enableXsrfCookie: true,
    methods: ['POST', 'PUT', 'PATCH', 'DELETE'],
  },
})

export default shieldConfig

enabled

Turn the CSRF guard on or off.

exceptRoutes

An array of route patterns to exempt from the CSRF protection. If your application has routes that accept form submissions via an API, you might want to exempt them.

For more advanced use cases, you may register a function to exempt specific routes dynamically.

{
  exceptRoutes: (ctx) => {
    // exempt all routes starting with /api/
    return ctx.request.url().includes('/api/')
  }
}

enableXsrfCookie

When enabled, Shield will store the CSRF token inside an encrypted cookie named XSRF-TOKEN, which can be read by the frontend JavaScript code.

This allows frontend request libraries like Axios to read the XSRF-TOKEN automatically and set it as a X-XSRF-TOKEN header when making Ajax requests without server-rendered forms.

You must keep the enableXsrfCookie disabled if you are not triggering Ajax requests programmatically.

methods

An array of HTTP methods to protect. All incoming requests for the mentioned methods must present a valid CSRF token.

cookieOptions

Configuration for the XSRF-TOKEN cookie. See cookies configuration for available options.

Defining CSP policy

CSP (Content security policy) protects your applications from XSS attacks by defining trusted sources for loading JavaScript, CSS, fonts, images, and so on.

The CSP guard is disabled by default. However, we recommend you enable it and configure the policy directives inside the config/shield.ts file.

import { defineConfig } from '@adonisjs/shield'

const shieldConfig = defineConfig({
  csp: {
    enabled: true,
    directives: {
      // policy directives go here
    },
    reportOnly: false,
  },
})

export default shieldConfig

enabled

Turn the CSP guard on or off.

directives

Configure the CSP directives. You can view the list of available directives on https://content-security-policy.com/

const shieldConfig = defineConfig({
  csp: {
    enabled: true,
    directives: {
      defaultSrc: [`'self'`],
      scriptSrc: [`'self'`, 'https://cdnjs.cloudflare.com'],
      fontSrc: [`'self'`, 'https://fonts.googleapis.com']
    },
    reportOnly: false,
  },
})

export default shieldConfig

reportOnly

The CSP policy will not block the resources when the reportOnly flag is enabled. Instead, it will report the violations on an endpoint configured using the reportUri directive.

const shieldConfig = defineConfig({
  csp: {
    enabled: true,
    directives: {
      defaultSrc: [`'self'`],
      reportUri: ['/csp-report']
    },
    reportOnly: true,
  },
})

Also, register the csp-report endpoint to collect the violation reports.

router.post('/csp-report', async ({ request }) => {
  const report = request.input('csp-report')
})

Using Nonce

You may allow inline script and style tags by defining the nonce attribute on them. The value of the nonce attribute can be accessed inside Edge templates using the cspNonce property.

<script nonce="{{ cspNonce }}">
  // Inline JavaScript
</script>
<style nonce="{{ cspNonce }}">
  /* Inline CSS */
</style>

Also, use the @nonce keyword inside the directives config to allow nonce-based inline scripts and styles.

const shieldConfig = defineConfig({
  csp: {
    directives: {
      defaultSrc: [`'self'`, '@nonce'],
    },
  },
})

Loading assets from the Vite Dev server

If you are using the Vite integration, you can use the following CSP keywords to allow assets served by the Vite Dev server.

The @viteDevUrl adds the Vite dev server URL to the allowed list.
The @viteHmrUrl adds the Vite HMR websocket server URL to the allowed list.

const shieldConfig = defineConfig({
  csp: {
    directives: {
      defaultSrc: [`'self'`, '@viteDevUrl'],
      connectSrc: ['@viteHmrUrl']
    },
  },
})

If you are deploying the Vite bundled output to a CDN server, you must replace @viteDevUrl with the @viteUrl keyword to allow assets from both the development server and the CDN server.

directives: {
  defaultSrc: [`'self'`, '@viteDevUrl'],
  defaultSrc: [`'self'`, '@viteUrl'],
  connectSrc: ['@viteHmrUrl']
},

Adding Nonce to styles injected by Vite

Currently, Vite does not allow defining a nonce attribute to the style tags injected by it inside the DOM. There is an open PR for the same, and we are hoping it will be resolved soon.

Configuring HSTS

The Strict-Transport-Security (HSTS) response header informs the browsers to always load the website using HTTPS.

You can configure the header directives using the config/shield.ts file.

import { defineConfig } from '@adonisjs/shield'

const shieldConfig = defineConfig({
  hsts: {
    enabled: true,
    maxAge: '180 days',
    includeSubDomains: true,
  },
})

enabled

Turn the hsts guard on or off.

maxAge

Defines the max-age attribute. The value should either be a number in seconds or a string-based time expression.

{
  // Remember for 10 seconds
  maxAge: 10,
}

{
  // Remember for 2 days
  maxAge: '2 days',
}

includeSubDomains

Defines the includeSubDomains directive to apply the setting on subdomains.

Configuring X-Frame protection

The X-Frame-Options header is used to indicate if a browser is allowed to render a website embedded inside an iframe, frame, embed, or object tags.

If you have configured CSP, you may instead use the frame-ancestors directive and disable the xFrame guard.

You can configure the header directives using the config/shield.ts file.

import { defineConfig } from '@adonisjs/shield'

const shieldConfig = defineConfig({
  xFrame: {
    enabled: true,
    action: 'DENY'
  },
})

enabled

Turn the xFrame guard on or off.

action

The action property defines the header value. It could be DENY, SAMEORIGIN, or ALLOW-FROM.

{
  action: 'DENY'
}

In the case of ALLOW-FROM, you must also define the domain property.

{
  action: 'ALLOW-FROM',
  domain: 'https://foo.com',
}

Disabling MIME sniffing

The X-Content-Type-Options header instructs browsers to follow the content-type header and not perform MIME sniffing by inspecting the content of an HTTP response.

Once you enable this guard, Shield will define the X-Content-Type-Options: nosniff header for all HTTP responses.

import { defineConfig } from '@adonisjs/shield'

const shieldConfig = defineConfig({
  contentTypeSniffing: {
    enabled: true,
  },
})

Security

CORS

Security

Rate limiting

Securing server-rendered applications

CSRF protection

Protecting forms

Handling exceptions

Config reference

Defining CSP policy

Using Nonce

Loading assets from the Vite Dev server

Adding Nonce to styles injected by Vite

Configuring HSTS

Configuring X-Frame protection

Disabling MIME sniffing

On this page