electron/docs/tutorial/sandbox.md

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

179 lines
7.7 KiB
Markdown
Raw Normal View History

# Process Sandboxing
One key security feature in Chromium is that processes can be executed within a sandbox.
The sandbox limits the harm that malicious code can cause by limiting access to most
system resources — sandboxed processes can only freely use CPU cycles and memory.
In order to perform operations requiring additional privilege, sandboxed processes
use dedicated communication channels to delegate tasks to more privileged processes.
In Chromium, sandboxing is applied to most processes other than the main process.
This includes renderer processes, as well as utility processes such as the audio service,
the GPU service and the network service.
See Chromium's [Sandbox design document][sandbox] for more information.
Starting from Electron 20, the sandbox is enabled for renderer processes without any
further configuration. If you want to disable the sandbox for a process, see the
[Disabling the sandbox for a single process](#disabling-the-sandbox-for-a-single-process)
section.
## Sandbox behaviour in Electron
Sandboxed processes in Electron behave _mostly_ in the same way as Chromium's do, but
Electron has a few additional concepts to consider because it interfaces with Node.js.
### Renderer processes
When renderer processes in Electron are sandboxed, they behave in the same way as a
regular Chrome renderer would. A sandboxed renderer won't have a Node.js
environment initialized.
Therefore, when the sandbox is enabled, renderer processes can only perform privileged
tasks (such as interacting with the filesystem, making changes to the system, or spawning
subprocesses) by delegating these tasks to the main process via inter-process
communication (IPC).
:::note
For more info on inter-process communication, check out our [IPC guide](./ipc.md).
:::
### Preload scripts
In order to allow renderer processes to communicate with the main process, preload
scripts attached to sandboxed renderers will still have a polyfilled subset of Node.js
APIs available. A `require` function similar to Node's `require` module is exposed,
but can only import a subset of Electron and Node's built-in modules:
* `electron` (following renderer process modules: `contextBridge`, `crashReporter`, `ipcRenderer`, `nativeImage`, `webFrame`, `webUtils`)
* [`events`](https://nodejs.org/api/events.html)
* [`timers`](https://nodejs.org/api/timers.html)
* [`url`](https://nodejs.org/api/url.html)
[node: imports](https://nodejs.org/api/esm.html#node-imports) are supported as well:
* [`node:events`](https://nodejs.org/api/events.html)
* [`node:timers`](https://nodejs.org/api/timers.html)
* [`node:url`](https://nodejs.org/api/url.html)
In addition, the preload script also polyfills certain Node.js primitives as globals:
* [`Buffer`](https://nodejs.org/api/buffer.html)
* [`process`](../api/process.md)
* [`clearImmediate`](https://nodejs.org/api/timers.html#timers_clearimmediate_immediate)
* [`setImmediate`](https://nodejs.org/api/timers.html#timers_setimmediate_callback_args)
Because the `require` function is a polyfill with limited functionality, you will not be
able to use [CommonJS modules][commonjs] to separate your preload script into multiple
files. If you need to split your preload code, use a bundler such as [webpack][webpack]
or [Parcel][parcel].
Note that because the environment presented to the `preload` script is substantially
more privileged than that of a sandboxed renderer, it is still possible to leak
privileged APIs to untrusted code running in the renderer process unless
[`contextIsolation`][context-isolation] is enabled.
## Configuring the sandbox
For most apps, sandboxing is the best choice. In certain use cases that are incompatible with
the sandbox (for instance, when using native node modules in the renderer),
it is possible to disable the sandbox for specific processes. This comes with security
risks, especially if any untrusted code or content is present in the unsandboxed process.
### Disabling the sandbox for a single process
In Electron, renderer sandboxing can be disabled on a per-process basis with
the `sandbox: false` preference in the [`BrowserWindow`][browser-window] constructor.
```js title='main.js'
app.whenReady().then(() => {
const win = new BrowserWindow({
webPreferences: {
sandbox: false
}
})
win.loadURL('https://google.com')
})
```
Sandboxing is also disabled whenever Node.js integration is enabled in the renderer.
This can be done through the BrowserWindow constructor with the `nodeIntegration: true` flag.
```js title='main.js'
app.whenReady().then(() => {
const win = new BrowserWindow({
webPreferences: {
nodeIntegration: true
}
})
win.loadURL('https://google.com')
})
```
### Enabling the sandbox globally
If you want to force sandboxing for all renderers, you can also use the
[`app.enableSandbox`][enable-sandbox] API. Note that this API has to be called before the
app's `ready` event.
```js title='main.js'
app.enableSandbox()
app.whenReady().then(() => {
// any sandbox:false calls are overridden since `app.enableSandbox()` was called.
const win = new BrowserWindow()
win.loadURL('https://google.com')
})
```
### Disabling Chromium's sandbox (testing only)
You can also disable Chromium's sandbox entirely with the [`--no-sandbox`][no-sandbox]
CLI flag, which will disable the sandbox for all processes (including utility processes).
We highly recommend that you only use this flag for testing purposes, and **never**
in production.
Note that the `sandbox: true` option will still disable the renderer's Node.js
environment.
## A note on rendering untrusted content
Rendering untrusted content in Electron is still somewhat uncharted territory,
though some apps are finding success (e.g. [Beaker Browser][beaker]).
Our goal is to get as close to Chrome as we can in terms of the security of
sandboxed content, but ultimately we will always be behind due to a few fundamental
issues:
1. We do not have the dedicated resources or expertise that Chromium has to
apply to the security of its product. We do our best to make use of what we
have, to inherit everything we can from Chromium, and to respond quickly to
security issues, but Electron cannot be as secure as Chromium without the
resources that Chromium is able to dedicate.
1. Some security features in Chrome (such as Safe Browsing and Certificate
Transparency) require a centralized authority and dedicated servers, both of
which run counter to the goals of the Electron project. As such, we disable
those features in Electron, at the cost of the associated security they
would otherwise bring.
1. There is only one Chromium, whereas there are many thousands of apps built
on Electron, all of which behave slightly differently. Accounting for those
differences can yield a huge possibility space, and make it challenging to
ensure the security of the platform in unusual use cases.
1. We can't push security updates to users directly, so we rely on app vendors
to upgrade the version of Electron underlying their app in order for
security updates to reach users.
While we make our best effort to backport Chromium security fixes to older
versions of Electron, we do not make a guarantee that every fix will be
backported. Your best chance at staying secure is to be on the latest stable
version of Electron.
[sandbox]: https://chromium.googlesource.com/chromium/src/+/main/docs/design/sandbox.md
[browser-window]: ../api/browser-window.md
[enable-sandbox]: ../api/app.md#appenablesandbox
[no-sandbox]: ../api/command-line-switches.md#--no-sandbox
[commonjs]: https://nodejs.org/api/modules.html#modules_modules_commonjs_modules
[webpack]: https://webpack.js.org/
[parcel]: https://parceljs.org/
[context-isolation]: ./context-isolation.md
[beaker]: https://github.com/beakerbrowser/beaker