---
url: /api/accounts.md
---
# Accounts
## Accounts-base {#accounts-base}
The Meteor Accounts system builds on top of the `userId` support in
[`publish`](./meteor#Subscription-userId) and [`methods`](./meteor#methods-userId). The core
packages add the concept of user documents stored in the database, and
additional packages add [secure password authentication](#passwords),
[integration with third party login services](#Meteor-loginWith%3CExternalService%3E),
and a [pre-built userinterface](/packages/accounts-ui.html).
The basic Accounts system is in the `accounts-base` package, but
applications typically include this automatically by adding one of the
login provider packages: `accounts-password`, `accounts-facebook`,
`accounts-github`, `accounts-google`, `accounts-meetup`,
`accounts-twitter`, or `accounts-weibo`.
Read more about customizing user accounts in the [Accounts](http://guide.meteor.com/accounts.html) article in the Meteor Guide.
### Accounts with Session Storage {#accounts-session-storage}
By default, Meteor uses Local Storage to store, among other things, login tokens in your browser session. But, for some applications, it makes sense to use Session Storage instead. Session Storage will not persist across client sessions. You can achieve this by adding this to your settings:
```json
{
// ... all other settings,
"public": {
// ... all your public settings
"packages": {
"accounts": {
"clientStorage": "session"
}
}
}
}
```
Retrieves the user record for the current user from
the [`Meteor.users`](#Meteor-users) collection.
On the client, the available fields will be those that
are published from the server (other fields won't be available on the
client). By default the server publishes `username`, `emails`, and
`profile` (writable by user). See [`Meteor.users`](#Meteor-users) for more on
the fields used in user documents.
On the server, this will fetch the record from the database. To improve the
latency of a method that uses the user document multiple times, save the
returned record to a variable instead of re-calling `Meteor.user()`.
Fetching the full user document can cause unnecessary database usage on the
server and over-reactivity on the client, particularly if you store lots of
custom data on it. Therefore it is recommended to use the `options`
parameter to only fetch the fields you need:
```js
import { Meteor } from "meteor/meteor";
const userName = Meteor.user({ fields: { "profile.name": 1 } }).profile.name;
```
Same as [`Meteor.user`](#Meteor-user), but returns a promise and is available on the server.
```js
import { Meteor } from "meteor/meteor";
const user = await Meteor.userAsync();
```
This collection contains one document per registered user. Here's an example
user document:
```js
{
_id: 'QwkSmTCZiw5KDx3L6', // Meteor.userId()
username: 'cool_kid_13', // Unique name
emails: [
// Each email address can only belong to one user.
{ address: 'cool@example.com', verified: true },
{ address: 'another@different.com', verified: false }
],
createdAt: new Date('Wed Aug 21 2013 15:16:52 GMT-0700 (PDT)'),
profile: {
// The profile is writable by the user by default.
name: 'Joe Schmoe'
},
services: {
facebook: {
id: '709050', // Facebook ID
accessToken: 'AAACCgdX7G2...AbV9AZDZD'
},
resume: {
loginTokens: [
{ token: '97e8c205-c7e4-47c9-9bea-8e2ccc0694cd',
when: 1349761684048 }
]
}
}
}
```
A user document can contain any data you want to store about a user. Meteor
treats the following fields specially:
- `username`: a unique String identifying the user.
- `emails`: an Array of Objects with keys `address` and `verified`;
an email address may belong to at most one user. `verified` is
a Boolean which is true if the user has [verified the address](#Accounts-verifyEmail) with a token sent over email.
- `createdAt`: the Date at which the user document was created.
- `profile`: an Object which the user can create and update with any data.
Do not store anything on `profile` that you wouldn't want the user to edit
unless you have a deny rule on the `Meteor.users` collection.
- `services`: an Object containing data used by particular
login services. For example, its `reset` field contains
tokens used by [forgot password](#Accounts-forgotPassword) links,
and its `resume` field contains tokens used to keep you
logged in between sessions.
Like all [Mongo.Collection](./collections.md)s, you can access all
documents on the server, but only those specifically published by the server are
available on the client. You can also use all Collection methods, for instance
`Meteor.users.remove` on the server to delete a user.
By default, the current user's `username`, `emails` and `profile` are
published to the client. You can publish additional fields for the
current user with:
::: code-group
```js [server.js]
Meteor.publish("userData", function () {
if (this.userId) {
return Meteor.users.find(
{ _id: this.userId },
{
fields: { other: 1, things: 1 },
}
);
} else {
this.ready();
}
});
```
```js [client.js]
Meteor.subscribe("userData");
```
:::
If the autopublish package is installed, information about all users
on the system is published to all clients. This includes `username`,
`profile`, and any fields in `services` that are meant to be public
(eg `services.facebook.id`,
`services.twitter.screenName`). Additionally, when using autopublish
more information is published for the currently logged in user,
including access tokens. This allows making API calls directly from
the client for services that allow this.
Users are by default allowed to specify their own `profile` field with
[`Accounts.createUser`](#Accounts-createUser) and modify it with
`Meteor.users.update`. To allow users to edit additional fields, use
[`Meteor.users.allow`](./collections.md#Mongo-Collection-allow). To forbid users from making any modifications to
their user document:
```js
import { Meteor } from "meteor/meteor";
Meteor.users.deny({ update: () => true });
```
For example, [the `accounts-ui` package](../packages/accounts-ui.md) uses this to display an
animation while the login request is being processed.
For example, when called in a user's browser, connections in that browser
remain logged in, but any other browsers or DDP clients logged in as that user
will be logged out.
If there are multiple users with a username or email only differing in case, a case sensitive match is required. Although `createUser` won't let you create users with ambiguous usernames or emails, this could happen with existing databases or if you modify the users collection directly.
This method can fail throwing one of the following errors:
- "Unrecognized options for login request [400]" if `user` or `password` is undefined.
- "Match failed [400]" if `user` isn't an Object or String, or `password` isn't a String.
- "User not found [403]" if the email or username provided in `user` doesn't belong to a registered user.
- "Incorrect password [403]" if the password provided is incorrect.
- "User has no password set [403]" if `user` doesn't have a password.
This function is provided by the `accounts-password` package. See the
[Passwords](#passwords) section below.
Logs the user in using a valid Meteor login token (also called a resume token). This is typically used to restore a user's session across browser reloads, between tabs, or across DDP connections (such as in multi-server setups).
**Arguments:**
- `token` (`String`): The login token to use for authentication. Usually obtained from `Accounts._storedLoginToken()` or from a previous login session.
- `callback` (`Function`, optional): Called with no arguments on success, or with a single `Error` argument on failure.
**Returns:**
- `void`
**Usage example:**
```js
import { Meteor } from "meteor/meteor";
const token = Accounts._storedLoginToken();
Meteor.loginWithToken(token, (error) => {
if (error) {
console.error("Login with token failed", error);
} else {
console.log("Logged in with token!");
}
});
```
**Notes:**
- If the token is invalid, expired, or revoked, the callback will be called with an error and the user will not be logged in.
- This method is used internally by Meteor to automatically restore login state on page reload and across tabs.
- Can be used with custom DDP connections to authenticate across multiple Meteor servers sharing the same database.
Available functions are:
- `Meteor.loginWithMeteorDeveloperAccount`
- `Meteor.loginWithFacebook`
- `options` may also include [Facebook's `auth_type` parameter](https://developers.facebook.com/docs/facebook-login/manually-build-a-login-flow#reaskperms)
- `Meteor.loginWithGithub`
- `Meteor.loginWithGoogle`
- `options` may also include [Google's additional URI parameters](https://developers.google.com/identity/protocols/OpenIDConnect#authenticationuriparameters)
- `Meteor.loginWithMeetup`
- `Meteor.loginWithTwitter`
- `options` may also include [Twitter's `force_login` parameter](https://dev.twitter.com/oauth/reference/get/oauth/authenticate)
- `Meteor.loginWithWeibo`
These functions initiate the login process with an external
service (eg: Facebook, Google, etc), using OAuth. When called they open a new pop-up
window that loads the provider's login page. Once the user has logged in
with the provider, the pop-up window is closed and the Meteor client
logs in to the Meteor server with the information provided by the external
service.
Requesting Permissions
In addition to identifying the user to your application, some services
have APIs that allow you to take action on behalf of the user. To
request specific permissions from the user, pass the
`requestPermissions` option the login function. This will cause the user
to be presented with an additional page in the pop-up dialog to permit
access to their data. The user's `accessToken` — with permissions
to access the service's API — is stored in the `services` field of
the user document. The supported values for `requestPermissions` differ
for each login service and are documented on their respective developer
sites:
- Facebook:
- GitHub:
- Google:
- Meetup:
- Twitter, Weibo, Meteor developer accounts: `requestPermissions` currently not supported
External login services typically require registering and configuring
your application before use. The easiest way to do this is with the
[`accounts-ui` package](../packages/accounts-ui.md) which presents a step-by-step guide
to configuring each service. However, the data can be also be entered
manually in the `ServiceConfiguration.configurations` collection, which
is exported by the `service-configuration` package.
## Configuring Services {#service-configuration}
First, add the service configuration package:
```bash
meteor add service-configuration
```
Then, inside the server of your app (this example is for the Weebo service), import `ServiceConfiguration`:
```js
import { ServiceConfiguration } from "meteor/service-configuration";
ServiceConfiguration.configurations.upsertAsync(
{ service: "weibo" },
{
$set: {
loginStyle: "popup",
clientId: "1292962797", // See table below for correct property name!
secret: "75a730b58f5691de5522789070c319bc",
},
}
);
```
Since Meteor 2.7 you no longer need to manually set the configuration and instead can use Meteor settings by setting your services under `Meteor.settings.packages.service-configuration.`. All the properties can be set under the service and will be added to the database as is, so make sure that they are correct. For the example above, the settings would look like:
```json
{
"packages": {
"service-configuration": {
"weibo": {
"loginStyle": "popup",
"clientId": "1292962797",
"secret": "75a730b58f5691de5522789070c319bc"
}
}
}
}
```
The correct property name to use for the API identifier (i.e. `clientId` in the above example) depends on the login service being used, so be sure to use the correct one:
| Property Name | Services |
| ------------- | -------------------------------------------------------- |
| `appId` | Facebook |
| `clientId` | Github, Google, Meetup, Meteor Developer Accounts, Weibo |
| `consumerKey` | Twitter |
Additionally, each external service has its own login provider package and login function. For
example, to support GitHub login, run the following in your terminal:
```bash
meteor add accounts-github
```
and use the `Meteor.loginWithGithub` function:
```js
import { Meteor } from "meteor/meteor";
Meteor.loginWithGithub(
{
requestPermissions: ["user", "public_repo"],
},
(error) => {
if (error) {
Session.set("errorMessage", error.reason || "Unknown error");
}
}
);
```
Login service configuration is sent from the server to the client over DDP when
your app starts up; you may not call the login function until the configuration
is loaded. The function `Accounts.loginServicesConfigured()` is a reactive data
source that will return true once the login service is configured; you should
not make login buttons visible or active until it is true.
Ensure that your [`$ROOT_URL`](./meteor.md#Meteor-absoluteUrl) matches the authorized
domain and callback URL that you configure with the external service (for
instance, if you are running Meteor behind a proxy server, `$ROOT_URL` should be
the externally-accessible URL, not the URL inside your proxy).
## Manual service configuration {#manual-service-configuration}
You can use `Accounts.loginServiceConfiguration` to view and edit the settings collection:
```js
import { Accounts } from "meteor/accounts-base";
Accounts.loginServiceConfiguration.find();
```
## Popup versus redirect flow {#popup-vs-redirect-flow}
When configuring OAuth login with a provider (such as Facebook or Google), Meteor lets you choose a popup- or redirect-based flow. In a popup-based flow, when a user logs in, they will be prompted to login at the provider in a popup window. In a redirect-based flow, the user's whole browser window will be redirected to the login provider, and the window will redirect back to your app when the login is completed.
You can also pick which type of login to do by passing an option to [`Meteor.loginWith`](#Meteor-loginWith%3CExternalService%3E)
Usually, the popup-based flow is preferable because the user will not have to reload your whole app at the end of the login flow. However, the popup-based flow requires browser features such as `window.close` and `window.opener` that are not available in all mobile environments. In particular, we recommend using `Meteor.loginWith({ loginStyle: 'redirect' })` in the following environments:
- Inside UIWebViews (when your app is loaded inside a mobile app)
- In Safari on iOS8 (`window.close` is not supported due to a bug)
Example:
```js
import { Accounts } from "meteor/accounts-base";
Accounts.ui.config({
requestPermissions: {
facebook: ["user_likes"],
github: ["user", "repo"],
},
requestOfflineToken: {
google: true,
},
passwordSignupFields: "USERNAME_AND_OPTIONAL_EMAIL",
});
```
Since Meteor 2.7 you can configure these in your Meteor settings under `Meteor.settings.public.packages.accounts-ui-unstyled`.
## Multi-server {#multi-server}
The `accounts-base` package exports two constructors, called
`AccountsClient` and `AccountsServer`, which are used to create the
`Accounts` object that is available on the client and the server,
respectively.
This predefined `Accounts` object (along with similar convenience methods
of `Meteor`, such as [`Meteor.logout`](#Meteor-logout)) is sufficient to
implement most accounts-related logic in Meteor apps. Nevertheless, these
two constructors can be instantiated more than once, to create multiple
independent connections between different accounts servers and their
clients, in more complicated authentication situations.
The `AccountsClient` and `AccountsServer` classes share a common
superclass, `AccountsCommon`. Methods defined on
`AccountsCommon.prototype` will be available on both the client and the
server, via the predefined `Accounts` object (most common) or any custom
`accountsClientOrServer` object created using the `AccountsClient` or
`AccountsServer` constructors (less common).
Here are a few of those methods:
From Meteor 2.5 you can set these in your Meteor settings under `Meteor.settings.packages.accounts-base`. Note that due to the nature of settings file you won't be able to set parameters that require functions.
See description of [AccountsCommon#onLoginFailure](#AccountsCommon-onLoginFailure)
for details.
Either the `onLogin` or the `onLoginFailure` callbacks will be called
for each login attempt. The `onLogin` callbacks are called after the
user has been successfully logged in. The `onLoginFailure` callbacks are
called after a login attempt is denied.
These functions return an object with a single method, `stop`. Calling
`stop()` unregisters the callback.
On the server, the callbacks get a single argument, the same attempt info
object as [`validateLoginAttempt`](#AccountsServer-validateLoginAttempt). On the
client, the callback argument is an object containing a single `error`
property set to the `Error`-object which was received from the failed login
attempt.
On the server, the `func` callback receives a single argument with the object below. On the
client, no arguments are passed.
```js
import { AccountsCommon } from "meteor/accounts-base";
const options = {
//...
};
const accountsCommon = new AccountsCommon(options);
accountsCommon.onLogout(({ user, connection, collection }) => {
console.log(user);
// ˆˆˆˆˆˆ The Meteor user object of the user which just logged out
console.log(connection);
// ˆˆˆˆˆˆ The connection object the request came in on. See
// `Meteor.onConnection` for details.
console.log(collection);
// ˆˆˆˆˆˆ The `collection` The name of the Mongo.Collection or the
// Mongo.Collection object to hold the users.
});
```
At most one of `options.connection` and `options.ddpUrl` should be
provided in any instantiation of `AccountsClient`. If neither is provided,
`Meteor.connection` will be used as the `.connection` property of the
`AccountsClient` instance.
Note that `AccountsClient` is currently available only on the client, due
to its use of browser APIs such as `window.localStorage`. In principle,
though, it might make sense to establish a client connection from one
server to another remote accounts server. Please [let us
know](https://github.com/meteor/meteor/wiki/Contributing-to-Meteor#feature-requests)
if you find yourself needing this server-to-server functionality.
These methods are defined on `AccountsClient.prototype`, and are thus
available only on the client:
These methods are defined on `AccountsServer.prototype`, and are thus
available only on the server:
This can be called multiple times. If any of the functions return `false` or
throw an error, the new user creation is aborted. To set a specific error
message (which will be displayed by [`accounts-ui`](../packages/accounts-ui.md)), throw a new
[`Meteor.Error`](./meteor#meteor-api).
Example:
```js
import { Accounts } from "meteor/accounts-base";
// Validate username, sending a specific error message on failure.
Accounts.validateNewUser((user) => {
if (user.username && user.username.length >= 3) {
return true;
} else {
throw new Meteor.Error(403, "Username must have at least 3 characters");
}
});
// Validate username, without a specific error message.
Accounts.validateNewUser((user) => {
return user.username !== "root";
});
```
If the user is being created as part of a login attempt from a client (eg,
calling [`Accounts.createUser`](#Accounts-createUser) from the client, or
[logging in for the first time with an external
service](#meteor_loginwithexternalservice)), these callbacks are called _before_
the [`Accounts.validateLoginAttempt`](#Accounts-validateLoginAttempt)
callbacks. If these callbacks succeed but those fail, the user will still be
created but the connection will not be logged in as that user.
Use this when you need to do more than simply accept or reject new user
creation. With this function you can programatically control the
contents of new user documents.
The function you pass will be called with two arguments: `options` and
`user`. The `options` argument comes
from [`Accounts.createUser`](#Accounts-createUser) for
password-based users or from an external service login flow. `options` may come
from an untrusted client so make sure to validate any values you read from
it. The `user` argument is created on the server and contains a
proposed user object with all the automatically generated fields
required for the user to log in, including a temporary `_id` (the final _id is
generated upon document insertion and not available in this function).
The function should return the user document (either the one passed in or a
newly-created object) with whatever modifications are desired. The returned
document is inserted directly into the [`Meteor.users`](#Meteor-users) collection.
The default create user function simply copies `options.profile` into
the new user document. Calling `onCreateUser` overrides the default
hook. This can only be called once.
Example:
```js
import { Accounts } from "meteor/accounts-base";
// Support for playing D&D: Roll 3d6 for dexterity.
Accounts.onCreateUser((options, user) => {
const customizedUser = Object.assign(
{
dexterity: _.random(1, 6) + _.random(1, 6) + _.random(1, 6),
},
user
);
// We still want the default hook's 'profile' behavior.
if (options.profile) {
customizedUser.profile = options.profile;
}
return customizedUser;
});
```
Call `validateLoginAttempt` with a callback to be called on login
attempts. It returns an object with a single method, `stop`. Calling
`stop()` unregisters the callback.
When a login attempt is made, the registered validate login callbacks
are called with a single argument, you can check the example:
```js
import { AccountsServer } from "meteor/accounts-base";
const options = {
//...
};
const accountsServer = new AccountsServer(options);
accountsServer.validateLoginAttempt(
({
type, // String
allowed, // Boolean
error, // Error
user, // Object
connection, // Object
collection, // Object
methodName, // String
methodArguments, // Array
}) => {
console.log(type);
// ˆˆˆˆˆˆ The service name, such as "password" or "twitter".
console.log(allowed);
// ˆˆˆˆˆˆ Whether this login is allowed and will be successful (if not aborted
// by any of the validateLoginAttempt callbacks). False if the login
// will not succeed (for example, an invalid password or the login was
// aborted by a previous validateLoginAttempt callback).
console.log(error);
// ˆˆˆˆˆˆ When `allowed` is false, the exception describing why the login
// failed. It will be a `Meteor.Error` for failures reported to the
// user (such as invalid password), and can be a another kind of
// exception for internal errors.
console.log(user);
// ˆˆˆˆˆˆ When it is known which user was attempting to login,
// the Meteor user object. This will always be present for successful logins.
console.log(connection);
// ˆˆˆˆˆˆ The `connection` object the request came in on. See
// [`Meteor.onConnection`](#meteor_onconnection) for details.
console.log(collection);
// ˆˆˆˆˆˆ The `collection` The name of the Mongo.Collection or the
// Mongo.Collection object to hold the users.
console.log(methodName);
// ˆˆˆˆˆˆ The name of the Meteor method being used to login.
// For example, "login", "loginWithPassword", or "loginWith".
console.log(methodArguments);
// ˆˆˆˆˆˆ An array of the arguments passed to the login method.
// For example, `["username", "password"]`
}
);
```
A validate login callback must return a truthy value for the login to
proceed. If the callback returns a falsy value or throws an
exception, the login is aborted. Throwing a `Meteor.Error` will
report the error reason to the user.
All registered validate login callbacks are called, even if one of the callbacks
aborts the login. The later callbacks will see the `allowed` field set to
`false` since the login will now not be successful. This allows later callbacks
to override an error from a previous callback; for example, you could override
the "Incorrect password" error with a different message.
Validate login callbacks that aren't explicitly trying to override a previous
error generally have no need to run if the attempt has already been determined
to fail, and should start with
```js
if (!attempt.allowed) {
return false;
}
```
Use this hook if you need to validate that user from an external service should
be allowed to login or create account.
```js
import { AccountsServer } from "meteor/accounts-base";
const options = {
//...
};
const accountsServer = new AccountsServer(options);
accountsServer.beforeExternalLogin(({ type, data, user }) => {
console.log(type);
// ˆˆˆˆˆˆ The service name, such as "google" or "twitter". Is a String
console.log(data);
// ˆˆˆˆˆˆ Data retrieved from the service (eg: email, name, etc)
// Is an Object.
console.log(user);
// ˆˆˆˆˆˆ If user was found in the database that matches the criteria from the service,
// their data will be provided here. Is an Object.
});
```
You should return a `Boolean` value, `true` if the login/registration should
proceed or `false` if it should terminate. In case of termination
the login attempt will throw an error `403`, with the message: `Login forbidden`.
When allowing your users to authenticate with an external service, the process will
eventually call `Accounts.updateOrCreateUserFromExternalService`. By default, this
will search for a user with the `service..id`, and if not found will
create a new user. As that is not always desirable, you can use this hook as an
escape hatch to look up a user with a different selector, probably by `emails.address` or `username`. Note the function will only be called if no user was found with the
`service..id` selector.
The function will be called with a single argument, the info object:
```js
import { AccountsServer } from "meteor/accounts-base";
const options = {
//...
};
const accountsServer = new AccountsServer(options);
accountsServer.setAdditionalFindUserOnExternalLogin(
({ serviceName, serviceData, options }) => {
// serviceName: String
// The external service name, such as "google" or "twitter".
// serviceData: Object
// The data returned by the service oauth request.
// options: Object
// An optional arugment passed down from the oauth service that may contain
// additional user profile information. As the data in `options` comes from an
// external source, make sure you validate any values you read from it.
}
);
```
The function should return either a user document or `undefined`. Returning a user
will result in the populating the `service.` in your user document,
while returning `undefined` will result in a new user account being created.
If you would prefer that a new account not be created, you could throw an error
instead of returning.
Example:
```js
// If a user has already been created, and used their Google email, this will
// allow them to sign in with the Meteor.loginWithGoogle method later, without
// creating a new user.
Accounts.setAdditionalFindUserOnExternalLogin(
({ serviceName, serviceData }) => {
if (serviceName === "google") {
// Note: Consider security implications. If someone other than the owner
// gains access to the account on the third-party service they could use
// the e-mail set there to access the account on your app.
// Most often this is not an issue, but as a developer you should be aware
// of how bad actors could play.
return Accounts.findUserByEmail(serviceData.email);
}
}
);
```
Use this to register your own custom authentication method. This is also used by all of the other inbuilt accounts packages to integrate with the accounts system.
There can be multiple login handlers that are registered. When a login request is made, it will go through all these handlers to find its own handler.
The registered handler callback is called with a single argument, the `options` object which comes from the login method. For example, if you want to login with a plaintext password, `options` could be `{ user: { username: }, password: }`,or `{ user: { email: }, password: }`.
The login handler should return `undefined` if it's not going to handle the login request or else the login result object.
Rate Limiting
By default, there are rules added to the [`DDPRateLimiter`](./DDPRateLimiter.md)
that rate limit logins, new user registration and password reset calls to a
limit of 5 requests per 10 seconds per session. These are a basic solution
to dictionary attacks where a malicious user attempts to guess the passwords
of legitimate users by attempting all possible passwords.
These rate limiting rules can be removed by calling
`Accounts.removeDefaultRateLimit()`. Please see the
[`DDPRateLimiter`](./DDPRateLimiter.md) docs for more information.
## Passwords {#passwords}
The `accounts-password` package contains a full system for password-based
authentication. In addition to the basic username and password-based
sign-in process, it also supports email-based sign-in including
address verification and password recovery emails.
### Password encryption and security
Starting from `accounts-passwords:4.0.0`, you can choose which algorithm is used by the Meteor server to store passwords : either [bcrypt](http://en.wikipedia.org/wiki/Bcrypt) or
[Argon2](http://en.wikipedia.org/wiki/Argon2) algorithm. Both are robust and contribute to
protect against embarrassing password leaks if the server's database is
compromised.
Before version 4.0.0, `bcrypt` was the only available option. argon2 has been introduced because it is considered the most secure option. This algorithm is specifically designed to resist GPU-based brute force attacks. For more details, see the [OWASP Password Storage Cheat Sheet](https://cheatsheetseries.owasp.org/cheatsheets/Password_Storage_Cheat_Sheet.html).
As of January 2025, **`bcrypt` is still the default option** to enable a smooth transition. In the future, `argon2` will replace `bcrypt` as default and `bcrypt` option will be deprecated.
Passwords are hashed on the client using **SHA-256** algorithm before being sent to the server. This ensures that sensitive data is never transmitted in plain text. Once received by the server, the hashed value is further encrypted and securely stored in the `Meteor.users` collection.
**About the migration process from `bcrypt` to `argon2`**
The transition from `bcrypt` to `argon2` happens automatically upon user login. If Argon2 encryption is enabled in an existing application, each user's password is re-encrypted during their next successful login.
- Step 1: The password is first validated against the existing `bcrypt` hash.
- Step 2: If authentication succeeds, the password is re-encrypted using `Argon2`.
- Step 3: The new `Argon2` hash replaces the old `bcrypt` hash in the database.
To monitor the migration progress, you can count users still using bcrypt:
```js
const bcryptUsers = await Meteor.users.find({ "services.password.bcrypt": { $exists: true } }).countAsync();
const totalUsers = await Meteor.users.find({ "services.password": { $exists: true } }).countAsync();
console.log("Remaining users to migrate:", bcryptUsers, "/", totalUsers);
```
Once `bcryptUsers` reaches 0, the migration is complete.
**Enabling Argon2 encryption**
To enable Argon2 encryption, you need a small configuration change on the server:
```js
Accounts.config({
argon2Enabled: true,
});
```
**Configuring `argon2` parameters**
One enabled, the `accounts-password` package allows customization of Argon2's parameters. The configurable options include:
- `type`: `argon2id` (provides a blend of resistance against GPU and side-channel attacks)
- `timeCost` (default: 2) – This controls the computational cost of the hashing process, affecting both the security level and performance.
- `memoryCost`: 19456 (19 MiB) - The amount of memory used by the algorithm in KiB per thread
- `parallelism`: 1 - The number of threads used by the algorithm
To update the values, use the following configuration:
```js
Accounts.config({
argon2Enabled: true,
argon2Type: "argon2id",
argon2TimeCost: 2,
argon2MemoryCost: 19456,
argon2Parallelism: 1,
});
```
Other Argon2 parameters, such as `hashLength`, are kept to default values:
- `hashLength`: 32 bytes - The length of the hash output in bytes
The default values are the minimum [OWASP recommendations for Argon2 parameters](https://cheatsheetseries.owasp.org/cheatsheets/Password_Storage_Cheat_Sheet.html#introduction). When updating these values, consider the trade-offs between security and performance on the target infrastructure.
For more information about Argon2's parameters, refer to the [argon2 options documentation](https://github.com/ranisalt/node-argon2/wiki/Options).
### Using passwords
To add password support to your application, run this command in your terminal:
```bash
meteor add accounts-password
```
> In addition to configuring the [`email`](./email.md) package's `MAIL_URL`, it is critical that you set proper values (specifically the `from` address) in [`Accounts.emailTemplates`](#Accounts-emailTemplates) to ensure proper delivery of e-mails!
You can construct your own user interface using the
functions below, or use the [`accounts-ui` package](../packages/accounts-ui.md) to
include a turn-key user interface for password-based sign-in.
On the client, this function logs in as the newly created user on
successful completion. On the server, it returns the newly created user
id.
On the client, you must pass `password` and at least one of `username` or `email` — enough information for the user to be able to log in again later. If there are existing users with a username or email only differing in case, `createUser` will fail. The callback's `error.reason` will be `'Username already exists.'` or `'Email already exists.'` In the latter case, the user can then either [login](accounts.html#Meteor-loginWithPassword) or [reset their password](#Accounts-resetPassword).
On the server, you do not need to specify `password`, but the user will not be able to log in until it has a password (eg, set with [`Accounts.setPasswordAsync`](#Accounts-setPasswordAsync)). To create an account without a password on the server and still let the user pick their own password, call `createUser` with the `email` option and then call [`Accounts.sendEnrollmentEmail`](#Accounts-sendEnrollmentEmail). This will send the user an email with a link to set their initial password.
By default the `profile` option is added directly to the new user document. To
override this behavior, use [`Accounts.onCreateUser`](#Accounts-onCreateUser).
This function is only used for creating users with passwords. The external
service login flows do not use this function.
Instead of modifying documents in the [`Meteor.users`](#Meteor-users) collection
directly, use these convenience functions which correctly check for case
insensitive duplicates before updates.
By default, an email address is added with `{ verified: false }`. Use
[`Accounts.sendVerificationEmail`](#Accounts-sendVerificationEmail) to send an
email with a link the user can use to verify their email address.
If the user trying to verify the email has 2FA enabled, this error will be thrown:
- "Email verified, but user not logged in because 2FA is enabled [2fa-enabled]": No longer signing in the user automatically if the user has 2FA enabled.
This function accepts tokens passed into the callback registered with
[`Accounts.onEmailVerificationLink`](#Accounts-onEmailVerificationLink).
Use the below functions to initiate password changes or resets from the server
or the client.
This triggers a call
to [`Accounts.sendResetPasswordEmail`](#Accounts-sendResetPasswordEmail)
on the server. When the user visits the link in this email, the callback
registered with [`Accounts.onResetPasswordLink`](#Accounts-onResetPasswordLink)
will be called.
If you are using the [`accounts-ui` package](../packages/accounts-ui.md), this is handled
automatically. Otherwise, it is your responsibility to prompt the user for the
new password and call `resetPassword`.
This function accepts tokens passed into the callbacks registered with
[`AccountsClient#onResetPasswordLink`](#Accounts-onResetPasswordLink) and
[`Accounts.onEnrollmentLink`](#Accounts-onEnrollmentLink).
If the user trying to reset the password has 2FA enabled, this error will be thrown:
- "Changed password, but user not logged in because 2FA is enabled [2fa-enabled]": No longer signing in the user automatically if the user has 2FA enabled.
When the user visits the link in this email, the callback registered with
[`AccountsClient#onResetPasswordLink`](#Accounts-onResetPasswordLink) will be called.
To customize the contents of the email, see
[`Accounts.emailTemplates`](#Accounts-emailTemplates).
When the user visits the link in this email, the callback registered with
[`Accounts.onEnrollmentLink`](#Accounts-onEnrollmentLink) will be called.
To customize the contents of the email, see
[`Accounts.emailTemplates`](#Accounts-emailTemplates).
When the user visits the link in this email, the callback registered with
[`Accounts.onEmailVerificationLink`](#Accounts-onEmailVerificationLink) will
be called.
To customize the contents of the email, see
[`Accounts.emailTemplates`](#Accounts-emailTemplates).
This is an `Object` with several fields that are used to generate text/html
for the emails sent by `sendResetPasswordEmail`, `sendEnrollmentEmail`,
and `sendVerificationEmail`.
Set the fields of the object by assigning to them:
- `from`: (**required**) A `String` with an [RFC5322](http://tools.ietf.org/html/rfc5322) From
address. By default, the email is sent from `no-reply@example.com`. **If you
want e-mails to send correctly, this should be changed to your own domain
as most e-mail providers will reject mail sent from `example.com`.**
- `siteName`: The public name of your application. Defaults to the DNS name of
the application (eg: `awesome.meteor.com`).
- `headers`: An `Object` for custom email headers as described in
[`Email.send`](./email.md#Email-send).
- `resetPassword`: An `Object` with the fields:
- `from`: A `Function` used to override the `from` address defined
by the `emailTemplates.from` field.
- `subject`: A `Function` that takes a user object and returns
a `String` for the subject line of a reset password email.
- `text`: An optional `Function` that takes a user object and a url, and
returns the body text for a reset password email.
- `html`: An optional `Function` that takes a user object and a
url, and returns the body html for a reset password email.
- `enrollAccount`: Same as `resetPassword`, but for initial password setup for
new accounts.
- `verifyEmail`: Same as `resetPassword`, but for verifying the users email
address.
Example:
```js
import { Accounts } from "meteor/accounts-base";
Accounts.emailTemplates.siteName = "AwesomeSite";
Accounts.emailTemplates.from = "AwesomeSite Admin ";
Accounts.emailTemplates.enrollAccount.subject = (user) => {
return `Welcome to Awesome Town, ${user.profile.name}`;
};
Accounts.emailTemplates.enrollAccount.text = (user, url) => {
return (
"You have been selected to participate in building a better future!" +
" To activate your account, simply click the link below:\n\n" +
url
);
};
Accounts.emailTemplates.resetPassword.from = () => {
// Overrides the value set in `Accounts.emailTemplates.from` when resetting
// passwords.
return "AwesomeSite Password Reset ";
};
Accounts.emailTemplates.verifyEmail = {
subject() {
return "Activate your account now!";
},
text(user, url) {
return `Hey ${user}! Verify your e-mail by following this link: ${url}`;
},
};
```
Enable 2FA for this package
You can add 2FA to your login flow by
using the package [accounts-2fa](../packages/accounts-2fa.md).
You can find an example showing how this would look like [here](../packages/accounts-2fa.md#working-with-accounts-password).
---
---
url: /packages/accounts-2fa.md
---
# accounts-2fa
This package allows you to provide a way for your users to enable 2FA on their accounts, using an authenticator app such as Google Authenticator, or 1Password. When the user is logged in on your app, they will be able to generate a new QR code and read this code on the app they prefer. After that, they'll start receiving their codes. Then, they can finish enabling 2FA on your app, and every time they try to log in to your app, you can redirect them to a place where they can provide a code they received from the authenticator.
To provide codes that are exactly compatible with all other Authenticator apps and services that implements TOTP, this package uses [node-2fa](https://www.npmjs.com/package/node-2fa) which works on top of [notp](https://github.com/guyht/notp), **that** implements TOTP ([RFC 6238](https://www.ietf.org/rfc/rfc6238.txt)) (the Authenticator standard), which is based on HOTP ([RFC 4226](https://www.ietf.org/rfc/rfc4226.txt)).
> This package is meant to be used with [`accounts-password`](../api/accounts.md#passwords) or [`accounts-passwordless`](./accounts-passwordless.md), so if you don't have either of those in your project, you'll need to add one of them. In the future, we want to enable the use of this package with other login methods, our oauth methods (Google, GitHub, etc...).
## 2FA Activation Flow {#activating-2fa}
The first step, in order to enable 2FA, is to generate a QR code so that the user can scan it in an authenticator app and start receiving codes.
Receives an `appName` which is the name of your app that will show up when the user scans the QR code. Also, a callback called, on success, with a QR code in SVG format, a QR secret, and the URI that can be used to activate the 2FA in an authenticator app,
or a single `Error` argument on failure.
On success, this function will also add an object to the logged user's services object containing the QR secret:
```js
services: {
...
twoFactorAuthentication: {
secret: "***"
}
}
```
Here it's an example on how to call this function:
```js
import { Buffer } from "buffer";
import { Accounts } from 'meteor/accounts-base';
// component
const [qrCode, setQrCode] = useState(null);
```
This method can fail throwing the following error:
- "The 2FA is activated. You need to disable the 2FA first before trying to generate a new activation code [2fa-activated]" if trying to generate an activation when the user already have 2FA enabled.
At this point, the 2FA won't be activated just yet. Now that the user has access to the codes generated by their authenticator app, you can call the function `Accounts.enableUser2fa`:
It should be called with a code that the users will receive from the authenticator app once they read the QR code. The callback is called with a single `Error` argument on failure. If the code provided is correct, a `type` will be added to the user's `twoFactorAuthentication` object and now 2FA is considered enabled:
```js
services: {
...
twoFactorAuthentication: {
type: "otp",
secret: "***",
}
}
```
To verify whether or not a user has 2FA enabled, you can call the function `Accounts.has2faEnabled`:
This function must be called when the user is logged in.
## Disabling 2FA {#disabling-2fa}
To disable 2FA for a user use this method:
To call this function the user must be already logged in.
## Log in with 2FA {#log-in-with-2fa}
Now that you have a way to allow your users to enable 2FA on their accounts, you can create a login flow based on that.
As said at the beginning of this guide, this package is currently working with two other packages: `accounts-password` and `accounts-passwordless`. Below there is an explanation on how to use this package with them.
## Working with accounts-password {#working-with-accounts-password}
When calling the function `Meteor.loginWithPassword`, if the 2FA is enabled for the user, an error will be returned to the callback, so you can redirect the user to a place where they can provide a code.
As an example:
```js
```
If the 2FA is not enabled, the user will be logged in normally.
The function you will need to call now to allow the user to login is `Meteor.loginWithPasswordAnd2faCode`:
Now you will be able to receive a code from the user and this function will verify if the code is valid. If it is, the user will be logged in.
So the call of this function should look something like this:
```js
```
This method can fail throwing one of the following errors:
- "2FA code must be informed [no-2fa-code]" if a 2FA code was not provided.
- "Invalid 2FA code [invalid-2fa-code]" if the provided 2FA code is invalid.
## Working with accounts-passwordless {#working-with-accounts-passwordless}
Following the same logic from the previous package, if the 2FA is enabled, an error will be returned to the callback of the function
[`Meteor.passwordlessLoginWithToken`](./accounts-passwordless.md#Meteor-passwordlessLoginWithToken),
then you can redirect the user to a place where they can provide a code.
Here is an example:
```js
```
Now you can call the function `Meteor.passwordlessLoginWithTokenAnd2faCode` that will allow you to provide a selector, token, and 2FA code:
This method can fail throwing one of the following errors:
- "2FA code must be informed [no-2fa-code]" if a 2FA code was not provided.
- "Invalid 2FA code [invalid-2fa-code]" if the provided 2FA code is invalid.
## Integrating an Authentication Package with accounts-2fa {#integrating-auth-package}
To integrate this package with any other existing Login method, it's necessary following two steps:
1 - For the client, create a new method from your current login method. So for example, from the method `Meteor.loginWithPassword` we created a new one called `Meteor.loginWithPasswordAnd2faCode`, and the only difference between them is that the latest one receives one additional parameter, the 2FA code, but we call the same function on the server side.
2 - For the server, inside the function that will log the user in, you verify if the function `Accounts._check2faEnabled` exists, and if yes, you call it providing the user object you want to check if the 2FA is enabled, and if either of these statements are false, you proceed with the login flow. This function exists only when the package `accounts-2fa` is added to the project.
If both statements are true, and the login validation succeeds, you verify if a code was provided: if not, throw an error; if it was provided, verify if the code is valid by calling the function `Accounts._isTokenValid`. if `Accounts._isTokenValid` returns false, throw an error.
Here it's an example:
```js
const result = validateLogin();
if (!result.error && Accounts._check2faEnabled?.(user)) {
if (!code) {
Accounts._handleError("2FA code must be informed.");
}
if (
!Accounts._isTokenValid(user.services.twoFactorAuthentication.secret, code)
) {
Accounts._handleError("Invalid 2FA code.");
}
}
return result;
```
---
---
url: /packages/accounts-ui.md
---
# accounts-ui
A turn-key user interface for Meteor Accounts.
To add Accounts and a set of login controls to an application, add the
`accounts-ui` package and at least one login provider package:
`accounts-password`, `accounts-facebook`, `accounts-github`,
`accounts-google`, `accounts-twitter`, or `accounts-weibo`.
Then simply add the `{{> loginButtons}}` helper to an HTML file. This
will place a login widget on the page. If there is only one provider configured
and it is an external service, this will add a login/logout button. If you use
`accounts-password` or use multiple external login services, this will add
a "Sign in" link which opens a dropdown menu with login options. If you plan to
position the login dropdown in the right edge of the screen, use
`{{> loginButtons align="right"}}` in order to get the dropdown to lay
itself out without expanding off the edge of the screen.
To configure the behavior of `{{> loginButtons}}`, use
[`Accounts.ui.config`](../api/accounts.md#loggingIn).
`accounts-ui` also includes modal popup dialogs to handle links from
[`sendResetPasswordEmail`](../api/accounts.md#Accounts-sendResetPasswordEmail),
[`sendVerificationEmail`](../api/accounts.md#Accounts-sendVerificationEmail),
and [`sendEnrollmentEmail`](../api/accounts.md#Accounts-sendEnrollmentEmail). These
do not have to be manually placed in HTML: they are automatically activated
when the URLs are loaded.
If you want to control the look and feel of your accounts system a little more, we recommend reading the [useraccounts](http://guide.meteor.com/accounts.html#useraccounts) section of the Meteor Guide.
---
---
url: /packages/appcache.md
---
# AppCache
> This package has been deprecated since [applicationCache](https://developer.mozilla.org/en-US/docs/Web/API/Window/applicationCache), which this package relies on, has been deprecated and is not available on the latest browsers. Plaese consider using [Service Worker](https://developer.mozilla.org/en-US/docs/Web/API/Service_Worker_API) as an replacement.
The `appcache` package stores the static parts of a Meteor application
(the client side Javascript, HTML, CSS, and images) in the browser's
[application cache](https://en.wikipedia.org/wiki/AppCache). To enable
caching simply add the `appcache` package to your project.
* Once a user has visited a Meteor application for the first time and
the application has been cached, on subsequent visits the web page
loads faster because the browser can load the application out of the
cache without contacting the server first.
* Hot code pushes are loaded by the browser in the background while the
app continues to run. Once the new code has been fully loaded the
browser is able to switch over to the new code quickly.
* The application cache allows the application to be loaded even when
the browser doesn't have an Internet connection, and so enables using
the app offline.
(Note however that the `appcache` package by itself doesn't make
*data* available offline: in an application loaded offline, a Meteor
Collection will appear to be empty in the client until the Internet
becomes available and the browser is able to establish a DDP
connection).
To turn AppCache off for specific browsers use:
```js
Meteor.AppCache.config({
chrome: false,
firefox: false
});
```
The supported browsers that can be enabled or disabled include, but are
not limited to, `android`, `chrome`, `chromium`, `chromeMobileIOS`,
`firefox`, `ie`, `mobileSafari` and `safari`.
Browsers limit the amount of data they will put in the application
cache, which can vary due to factors such as how much disk space is
free. Unfortunately if your application goes over the limit rather
than disabling the application cache altogether and running the
application online, the browser will instead fail that particular
*update* of the cache, leaving your users running old code.
Thus it's best to keep the size of the cache below 5MB. The
`appcache` package will print a warning on the Meteor server console
if the total size of the resources being cached is over 5MB.
Starting from `appcache@1.2.5`, if you need more advanced logic
to enable/disable the cache, you can use the `enableCallback` option
that is evaluated on a per-request basis. For example:
```js
// Enable offline mode using a value from database and certificate validation
Meteor.AppCache.config({
// This option is available starting from appcache@1.2.4
enableCallback: () => {
if (!getSettingsFromDb("public.appcache_enabled")) {
return false;
}
const validation = validateClientCert({
clientCertPayload: req.headers["x-client-cert"],
url: req.url.href,
});
return validation.passed;
},
});
```
If you have files too large to fit in the cache you can disable
caching by URL prefix. For example,
```js
Meteor.AppCache.config({ onlineOnly: ['/online/'] });
```
causes files in your `public/online` directory to not be cached, and
so they will only be available online. You can then move your large
files into that directory and refer to them at the new URL:
```html
```
If you'd prefer not to move your files, you can use the file names
themselves as the URL prefix:
```js
Meteor.AppCache.config({
onlineOnly: [
'/bigimage.jpg',
'/largedata.json'
]
});
```
though keep in mind that since the exclusion is by prefix (this is a
limitation of the application cache manifest), excluding
`/largedata.json` will also exclude such URLs as
`/largedata.json.orig` and `/largedata.json/file1`.
For more information about how Meteor interacts with the application
cache, see the
[AppCache page](https://github.com/meteor/meteor/wiki/AppCache)
in the Meteor wiki.
---
---
url: /tutorials/security/security.md
---
# Application Security for Production
After reading this tutorial, you'll know:
1. The security surface area of a Meteor app.
2. How to secure Meteor Methods, publications, and source code.
3. Where to store secret keys in development and production.
4. How to follow a security checklist when auditing your app.
5. How App Protection works in Galaxy Hosting.
[[toc]]
## Introduction
Securing a web application is all about understanding security domains and understanding the attack surface between these domains. In a Meteor app, things are pretty simple:
1. Code that runs on the server can be trusted.
2. Everything else: code that runs on the client, data sent through Method and publication arguments, etc, can't be trusted.
In practice, this means that you should do most of your security and validation on the boundary between these two domains. In simple terms:
1. Validate and check all inputs that come from the client.
2. Don't leak any secret information to the client.
### Concept: Attack surface
Since Meteor apps are often written in a style that puts client and server code together, it's extra important to be aware what is running on the client, what is running on the server, and what the boundaries are. Here's a complete list of places security checks need to be done in a Meteor app:
1. **Methods**: Any data that comes in through Method arguments needs to be validated, and Methods should not return data the user shouldn't have access to.
2. **Publications**: Any data that comes in through publication arguments needs to be validated, and publications should not return data the user shouldn't have access to.
3. **Served files**: You should make sure none of the source code or configuration files served to the client have secret data.
Each of these points will have their own section below.
#### Avoid allow/deny
In this guide, we're going to take a strong position that using [allow](/api/collections.html#Mongo-Collection-allow) or [deny](/api/collections.html#Mongo-Collection-deny) to run MongoDB queries directly from the client is not a good idea. The main reason is that it is hard to follow the principles outlined above. It's extremely difficult to validate the complete space of possible MongoDB operators, which could potentially grow over time with new versions of MongoDB.
There have been several articles about the potential pitfalls of accepting MongoDB update operators from the client, in particular the [Allow & Deny Security Challenge](https://web.archive.org/web/20220705130732/https://www.discovermeteor.com/blog/allow-deny-security-challenge/) and its [results](https://web.archive.org/web/20220819163744/https://www.discovermeteor.com/blog/allow-deny-challenge-results/), both on the Discover Meteor blog.
Given the points above, we recommend that all Meteor apps should use Methods to accept data input from the client, and restrict the arguments accepted by each Method as tightly as possible.
Here's a code snippet to add to your server code which disables client-side updates on a collection. This will make sure no other part of your app can use `allow` with the collection `Lists`:
```js
// Deny all client-side updates on the Lists collection
Lists.deny({
insert() { return true; },
update() { return true; },
remove() { return true; },
});
```
## Methods
Methods are the way your Meteor server accepts inputs and data from the outside world, so it's natural that they are the most important topic for security. If you don't properly secure your Methods, users can end up modifying your database in unexpected ways - editing other people's documents, deleting data, or messing up your database schema causing the app to crash.
### Validate all arguments
It's much easier to write clean code if you can assume your inputs are correct, so it's valuable to validate all Method arguments before running any actual business logic. You don't want someone to pass a data type you aren't expecting and cause unexpected behavior.
Consider that if you are writing unit tests for your Methods, you would need to test all possible kinds of input to the Method; validating the arguments restricts the space of inputs you need to unit test, reducing the amount of code you need to write overall. It also has the extra bonus of being self-documenting; someone else can come along and read the code to find out what kinds of parameters a Method is looking for.
Just as an example, here's a situation where not checking arguments can be disastrous:
```js
Meteor.methods({
removeWidget(id) {
if (!this.userId) {
throw new Meteor.Error('removeWidget.unauthorized');
}
Widgets.remove(id);
}
});
```
If someone comes along and passes a non-ID selector like `{}`, they will end up deleting the entire collection.
### jam:method
To help you write good Methods that exhaustively validate their arguments, you can use a community package for Methods that enforces argument validation. Read more about how to use it in the [documentation for jam:method](/community-packages/jam-method.html). The rest of the code samples in this article will assume that you are using this package. If you aren't, you can still apply the same principles but the code will look a little different.
### Never pass userId from the client
The `this` context inside every Meteor Method has some useful information about the current connection, and the most useful is [`this.userId`](/api/meteor.html#methods-userId). This property is managed by the DDP login system, and is guaranteed by the framework itself to be secure following widely-used best practices.
Given that the user ID of the current user is available through this context, you should never pass the ID of the current user as an argument to a Method. This would allow any client of your app to pass any user ID they want. Let's look at an example:
```js
// #1: Bad! The client could pass any user ID and set someone else's name
async run({ userId, newName }) {
await Meteor.users.updateAsync(userId, {
$set: { name: newName }
});
}
// #2: Good, the client can only set the name on the currently logged in user
async run({ newName }) {
await Meteor.users.updateAsync(this.userId, {
$set: { name: newName }
});
}
```
The _only_ times you should be passing any user ID as an argument are the following:
1. This is a Method only accessible by admin users, who are allowed to edit other users. See the section about [user roles](/packages/roles) to learn how to define roles and check that a user is in a certain role.
2. This Method doesn't modify the other user, but uses it as a target; for example, it could be a Method for sending a private message, or adding a user as a friend.
### One Method per action
The best way to make your app secure is to understand all of the possible inputs that could come from an untrusted source, and make sure that they are all handled correctly. The easiest way to understand what inputs can come from the client is to restrict them to as small of a space as possible. This means your Methods should all be specific actions, and shouldn't take a multitude of options that change the behavior in significant ways. The end goal is that you can look at each Method in your app and validate or test that it is secure. Here's a secure example Method from the Todos example app:
```js
export const makePrivate = createMethod({
name: 'lists.makePrivate',
schema: new SimpleSchema({
listId: { type: String }
}),
async run({ listId }) {
if (!this.userId) {
throw new Meteor.Error('lists.makePrivate.notLoggedIn',
'Must be logged in to make private lists.');
}
const list = await Lists.findOneAsync(listId);
if (list.isLastPublicList()) {
throw new Meteor.Error('lists.makePrivate.lastPublicList',
'Cannot make the last public list private.');
}
await Lists.updateAsync(listId, {
$set: { userId: this.userId }
});
Lists.userIdDenormalizer.set(listId, this.userId);
}
});
```
You can see that this Method does a _very specific thing_ - it makes a single list private. An alternative would have been to have a Method called `setPrivacy`, which could set the list to private or public, but it turns out that in this particular app the security considerations for the two related operations - `makePrivate` and `makePublic` - are very different. By splitting our operations into different Methods, we make each one much clearer. It's obvious from the above Method definition which arguments we accept, what security checks we perform, and what operations we do on the database.
However, this doesn't mean you can't have any flexibility in your Methods. Let's look at an example:
```js
Meteor.users.methods.setUserData = createMethod({
name: 'Meteor.users.methods.setUserData',
schema: new SimpleSchema({
fullName: { type: String, optional: true },
dateOfBirth: { type: Date, optional: true },
}),
async run(fieldsToSet) {
return (await Meteor.users.updateAsync(this.userId, {
$set: fieldsToSet
}));
}
});
```
The above Method is great because you can have the flexibility of having some optional fields and only passing the ones you want to change. In particular, what makes it possible for this Method is that the security considerations of setting one's full name and date of birth are the same - we don't have to do different security checks for different fields being set. Note that it's very important that the `$set` query on MongoDB is generated on the server - we should never take MongoDB operators as-is from the client, since they are hard to validate and could result in unexpected side effects.
#### Refactoring to reuse security rules
You might run into a situation where many Methods in your app have the same security checks. This can be simplified by factoring out the security into a separate module, wrapping the Method body, or extending the `Mongo.Collection` class to do security inside the `insert`, `update`, and `remove` implementations on the server. However, implementing your client-server communication via specific Methods is still a good idea rather than sending arbitrary `update` operators from the client, since a malicious client can't send an `update` operator that you didn't test for.
### Rate limiting
Like REST endpoints, Meteor Methods can be called from anywhere - a malicious program, script in the browser console, etc. It is easy to fire many Method calls in a very short amount of time. This means it can be easy for an attacker to test lots of different inputs to find one that works. Meteor has built-in rate limiting for password login to stop password brute-forcing, but it's up to you to define rate limits for your other Methods.
In the Todos example app, we use the following code to set a basic rate limit on all Methods:
```js
// Get list of all method names on Lists
const LISTS_METHODS = _.pluck([
insert,
makePublic,
makePrivate,
updateName,
remove,
], 'name');
// Only allow 5 list operations per connection per second
if (Meteor.isServer) {
DDPRateLimiter.addRule({
name(name) {
return _.contains(LISTS_METHODS, name);
},
// Rate limit per connection ID
connectionId() { return true; }
}, 5, 1000);
}
```
This will make every Method only callable 5 times per second per connection. This is a rate limit that shouldn't be noticeable by the user at all, but will prevent a malicious script from totally flooding the server with requests. You will need to tune the limit parameters to match your app's needs.
If you're using `jam:method`, it comes with built-in [rate-limiting](https://github.com/jamauro/method#rate-limiting).
## Publications
Publications are the primary way a Meteor server can make data available to a client. While with Methods the primary concern was making sure users can't modify the database in unexpected ways, with publications the main issue is filtering the data being returned so that a malicious user can't get access to data they aren't supposed to see.
### You can't do security at the rendering layer
In a server-side-rendered framework like Ruby on Rails, it's sufficient to not display sensitive data in the returned HTML response. In Meteor, since the rendering is done on the client, an `if` statement in your HTML template is not secure; you need to do security at the data level to make sure that data is never sent in the first place.
### Rules about Methods still apply
All of the points above about Methods apply to publications as well:
1. Validate all arguments using `check` or npm `simpl-schema`.
1. Never pass the current user ID as an argument.
1. Don't take generic arguments; make sure you know exactly what your publication is getting from the client.
1. Use rate limiting to stop people from spamming you with subscriptions.
### Always restrict fields
[`Mongo.Collection#find` has an option called `projection`](/api/collections.html#Mongo-Collection-find) which lets you filter the fields on the fetched documents. You should always use this in publications to make sure you don't accidentally publish secret fields.
For example, you could write a publication, then later add a secret field to the published collection. Now, the publication would be sending that secret to the client. If you filter the fields on every publication when you first write it, then adding another field won't automatically publish it.
```js
// #1: Bad! If we add a secret field to Lists later, the client
// will see it
Meteor.publish('lists.public', function () {
return Lists.find({userId: {$exists: false}});
});
// #2: Good, if we add a secret field to Lists later, the client
// will only publish it if we add it to the list of fields
Meteor.publish('lists.public', function () {
return Lists.find({userId: {$exists: false}}, {
projection: {
name: 1,
incompleteCount: 1,
userId: 1
}
});
});
```
If you find yourself repeating the fields often, it makes sense to factor out a dictionary of public fields that you can always filter by, like so:
```js
// In the file where Lists is defined
Lists.publicFields = {
name: 1,
incompleteCount: 1,
userId: 1
};
```
Now your code becomes a bit simpler:
```js
Meteor.publish('lists.public', function () {
return Lists.find({userId: {$exists: false}}, {
projection: Lists.publicFields
});
});
```
### Publications and userId
The data returned from publications will often be dependent on the currently logged in user, and perhaps some properties about that user - whether they are an admin, whether they own a certain document, etc.
Publications are not reactive, and they only re-run when the currently logged in `userId` changes, which can be accessed through `this.userId`. Because of this, it's easy to accidentally write a publication that is secure when it first runs, but doesn't respond to changes in the app environment. Let's look at an example:
```js
// #1: Bad! If the owner of the list changes, the old owner will still see it
Meteor.publish('list', async function (listId) {
check(listId, String);
const list = await Lists.findOneAsync(listId);
if (list.userId !== this.userId) {
throw new Meteor.Error('list.unauthorized',
'This list doesn\'t belong to you.');
}
return Lists.find(listId, {
projection: {
name: 1,
incompleteCount: 1,
userId: 1
}
});
});
// #2: Good! When the owner of the list changes, the old owner won't see it anymore
Meteor.publish('list', function (listId) {
check(listId, String);
return Lists.find({
_id: listId,
userId: this.userId
}, {
projection: {
name: 1,
incompleteCount: 1,
userId: 1
}
});
});
```
In the first example, if the `userId` property on the selected list changes, the query in the publication will still return the data, since the security check in the beginning will not re-run. In the second example, we have fixed this by putting the security check in the returned query itself.
Unfortunately, not all publications are as simple to secure as the example above. For more tips on how to use `reywood:publish-composite` to handle reactive changes in publications, see the [data loading article](https://guide.meteor.com/data-loading#complex-auth).
### Passing options
For certain applications, for example pagination, you'll want to pass options into the publication to control things like how many documents should be sent to the client. There are some extra considerations to keep in mind for this particular case.
1. **Passing a limit**: In the case where you are passing the `limit` option of the query from the client, make sure to set a maximum limit. Otherwise, a malicious client could request too many documents at once, which could raise performance issues.
2. **Passing in a filter**: If you want to pass fields to filter on because you don't want all of the data, for example in the case of a search query, make sure to use MongoDB `$and` to intersect the filter coming from the client with the documents that client should be allowed to see. Also, you should whitelist the keys that the client can use to filter - if the client can filter on secret data, it can run a search to find out what that data is.
3. **Passing in fields**: If you want the client to be able to decide which fields of the collection should be fetched, make sure to intersect that with the fields that client is allowed to see, so that you don't accidentally send secret data to the client.
In summary, you should make sure that any options passed from the client to a publication can only restrict the data being requested, rather than extending it.
## Served files
Publications are not the only place the client gets data from the server. The set of source code files and static assets that are served by your application server could also potentially contain sensitive data:
1. Business logic an attacker could analyze to find weak points.
1. Secret algorithms that a competitor could steal.
1. Secret API keys.
### Secret server code
While the client-side code of your application is necessarily accessible by the browser, every application will have some secret code on the server that you don't want to share with the world.
Secret business logic in your app should be located in code that is only loaded on the server. This means it is in a `server/` directory of your app, in a package that is only included on the server, or in a file inside a package that was loaded only on the server.
If you have a Meteor Method in your app that has secret business logic, you might want to split the Method into two functions - the optimistic UI part that will run on the client, and the secret part that runs on the server. Most of the time, putting the entire Method on the server doesn't result in the best user experience. Let's look at an example, where you have a secret algorithm for calculating someone's MMR (ranking) in a game:
```js
// In a server-only file, for example /imports/server/mmr.js
export const MMR = {
updateWithSecretAlgorithm(userId) {
// your secret code here
}
}
```
```js
// In a file loaded on client and server
Meteor.users.methods.updateMMR = new createMethod({
name: 'Meteor.users.methods.updateMMR',
validate: null,
run() {
if (this.isSimulation) {
// Simulation code for the client (optional)
} else {
const { MMR } = require('/imports/server/mmr.js');
MMR.updateWithSecretAlgorithm(this.userId);
}
}
});
```
::: warning
Note that while the Method is defined on the client, the actual secret logic is only accessible from the server and the code will **not** be included in the client bundle. Keep in mind that code inside `if (Meteor.isServer)` and `if (!this.isSimulation)` blocks is still sent to the client, it is just not executed. So don't put any secret code in there.
:::
Secret API keys should never be stored in your source code at all, the next section will talk about how to handle them.
## Securing API keys
Every app will have some secret API keys or passwords:
1. Your database password.
1. API keys for external APIs.
These should never be stored as part of your app's source code in version control, because developers might copy code around to unexpected places and forget that it contains secret keys. You can keep your keys separately in [Dropbox](https://www.dropbox.com/), [LastPass](https://lastpass.com), or another service, and then reference them when you need to deploy the app.
You can pass settings to your app through a _settings file_ or an _environment variable_. Most of your app settings should be in JSON files that you pass in when starting your app. You can start your app with a settings file by passing the `--settings` flag:
```sh
# Pass development settings when running your app locally
meteor --settings development.json
# Pass production settings when deploying your app to Galaxy
meteor deploy myapp.com --settings production.json
```
Here's what a settings file with some API keys might look like:
```js
{
"facebook": {
"appId": "12345",
"secret": "1234567"
}
}
```
In your app's JavaScript code, these settings can be accessed from the variable `Meteor.settings`.
[Read more about managing keys and settings in the Deployment article.](https://guide.meteor.com/deployment)
### Settings on the client
In most normal situations, API keys from your settings file will only be used by the server, and by default the data passed in through `--settings` is only available on the server. However, if you put data under a special key called `public`, it will be available on the client. You might want to do this if, for example, you need to make an API call from the client and are OK with users knowing that key. Public settings will be available on the client under `Meteor.settings.public`.
### Never store valuable information in public property in settings file
It's ok if you want to make some properties of your settings file accessible to the client but never put any valuable information inside the `public` property. Either explicity store it under `private` property or in its own `property`. Any property that's not under `public` is treated as private by default in Meteor.
```javascript
{
"public": {"publicKey": "xxxxx"},
"private": {"privateKey": "xxxxx"}
}
```
or
```javascript
{
"public": {"publicKey": "xxxxx"},
"privateKey": "xxxxx"
}
```
#### Example: API keys for OAuth
For the `accounts-facebook` package to pick up these keys, you need to add them to the service configuration collection in the database. Here's how you do that:
First, add the `service-configuration` package:
```sh
meteor add service-configuration
```
Then, upsert into the `ServiceConfiguration` collection using private settings:
```js
ServiceConfiguration.configurations.upsert({
service: "facebook"
}, {
$set: {
appId: Meteor.settings.private.facebook.appId,
loginStyle: "popup",
secret: Meteor.settings.private.facebook.secret
}
});
```
Now, `accounts-facebook` will be able to find that API key and Facebook login will work properly.
## SSL
This is a very short section, but it deserves its own place in the table of contents.
**Every production Meteor app that handles user data should run with SSL.**
Yes, Meteor does hash your password or login token on the client before sending it over the wire, but that only prevents an attacker from figuring out your password - it doesn't prevent them from logging in as you, since they could send the hashed password to the server to log in! No matter how you slice it, logging in requires the client to send sensitive data to the server, and the only way to secure that transfer is by using SSL. Note that the same issue is present when using cookies for authentication in a normal HTTP web application, so any app that needs to reliably identify users should be running on SSL.
#### Setting up SSL
* On [Galaxy](https://guide.meteor.com/deployment#galaxy), configuration of SSL is automatic. [See the help article about SSL on Galaxy](https://help.galaxycloud.app/en/article/encryption-pt8wbl/).
* If you are running on your own [infrastructure](https://guide.meteor.com/deployment#custom-deployment), there are a few options for setting up SSL, mostly through configuring a proxy web server. See the articles: [Josh Owens on SSL and Meteor](http://joshowens.me/ssl-and-meteor-js/), [SSL on Meteorpedia](http://www.meteorpedia.com/read/SSL), and [Digital Ocean tutorial with an Nginx config](https://www.digitalocean.com/community/tutorials/how-to-deploy-a-meteor-js-application-on-ubuntu-14-04-with-nginx).
#### Forcing SSL
Generally speaking, all production HTTP requests should go over HTTPS, and all WebSocket data should be sent over WSS.
It's best to handle the redirection from HTTP to HTTPS on the platform which handles the SSL certificates and termination.
* On [Galaxy](https://guide.meteor.com/deployment#galaxy), enable the "Force HTTPS" setting on a specific domain in the "Domains & Encryption" section of the application's "Settings" tab.
* Other deployments *may* have control panel options or may need to be manually configured on the proxy server (e.g. HAProxy, nginx, etc.). The articles linked above provide some assistance on this.
In the event that a platform does not offer the ability to configure this, the `force-ssl` package can be added to the project and Meteor will attempt to intelligently redirect based on the presence of the `x-forwarded-for` header.
## HTTP Headers
HTTP headers can be used to improve the security of apps, although these are not a silver bullet, they will assist users in mitigating more common attacks.
### Recommended: Helmet
Although there are many great open source solutions for setting HTTP headers, Meteor recommends [Helmet](https://helmetjs.github.io/). Helmet is a collection of 12 smaller middleware functions that set HTTP headers.
First, install helmet.
```js
meteor npm install helmet --save
```
By default, Helmet can be used to set various HTTP headers (see link above). These are a good starting point for mitigating common attacks. To use the default headers, users should use the following code anywhere in their server side meteor startup code.
> Note: Meteor has not extensively tested each header for compatibility with Meteor. Only headers listed below have been tested.
```js
// With other import statements
import helmet from "helmet";
// Within server side Meter.startup()
WebApp.handlers.use(helmet())
```
At a minimum, Meteor recommends users to set the following headers. Note that code examples shown below are specific to Helmet.
### Content Security Policy
> Note: Content Security Policy is not configured using Helmet's default header configuration.
From MDN, Content Security Policy (CSP) is an added layer of security that helps to detect and mitigate certain types of attacks, including Cross Site Scripting (XSS) and data injection attacks. These attacks are used for everything from data theft to site defacement or distribution of malware.
It is recommended that users use CSP to protect their apps from access by third parties. CSP assists to control how resources are loaded into your application.
By default, Meteor recommends unsafe inline scripts and styles are allowed, since many apps typically use them for analytics, etc. Unsafe eval is disallowed, and the only allowable content source is same origin or data, except for connect which allows anything (since meteor apps make websocket connections to a lot of different origins). Browsers will also be told not to sniff content types away from declared content types.
```js
// With other import statements
import helmet from "helmet";
// Within server side Meter.startup()
WebApp.handlers.use(
helmet.contentSecurityPolicy({
directives: {
defaultSrc: ["'self'"],
scriptSrc: ["'self'", "'unsafe-inline'"],
connectSrc: ["*"],
imgSrc: ["'self'"],
styleSrc: ["'self'", "'unsafe-inline'"],
}
})
);
```
Helmet supports a large number of directives, users should further customise their CSP based on their needs. For more detail please read the following guide: [Content Security Policy](https://helmetjs.github.io/docs/csp/). CSP can be complex, so in addition there are some excellent tools out there to help, including [Google's CSP Evaluator](https://csp-evaluator.withgoogle.com/), [Report-URI's CSP Builder](https://report-uri.com/home/generate), [CSP documentation](https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Content-Security-Policy/script-src) from Mozilla and [CSPValidator](https://cspvalidator.org/).
The following example presents a potential CSP and other Security Headers used in a Production Meteor Application.
This configuration may require customization, depending on your setup and use-cases.
```javascript
/* global __meteor_runtime_config__ */
import { Meteor } from 'meteor/meteor'
import { WebApp } from 'meteor/webapp'
import { Autoupdate } from 'meteor/autoupdate'
import { check } from 'meteor/check'
import crypto from 'crypto'
import helmet from 'helmet'
const self = '\'self\''
const data = 'data:'
const unsafeEval = '\'unsafe-eval\''
const unsafeInline = '\'unsafe-inline\''
const allowedOrigins = Meteor.settings.allowedOrigins
// create the default connect source for our current domain in
// a multi-protocol compatible way (http/ws or https/wss)
const url = Meteor.absoluteUrl()
const domain = url.replace(/http(s)*:\/\//, '').replace(/\/$/, '')
const s = url.match(/(?!=http)s(?=:\/\/)/) ? 's' : ''
const usesHttps = s.length > 0
const connectSrc = [
self,
`http${s}://${domain}`,
`ws${s}://${domain}`
]
// Prepare runtime config for generating the sha256 hash
// It is important, that the hash meets exactly the hash of the
// script in the client bundle.
// Otherwise the app would not be able to start, since the runtimeConfigScript
// is rejected __meteor_runtime_config__ is not available, causing
// a cascade of follow-up errors.
const runtimeConfig = Object.assign(__meteor_runtime_config__, Autoupdate, {
// the following lines may depend on, whether you called Accounts.config
// and whether your Meteor app is a "newer" version
accountsConfigCalled: true,
isModern: true
})
// add client versions to __meteor_runtime_config__
Object.keys(WebApp.clientPrograms).forEach(arch => {
__meteor_runtime_config__.versions[arch] = {
version: Autoupdate.autoupdateVersion || WebApp.clientPrograms[arch].version(),
versionRefreshable: Autoupdate.autoupdateVersion || WebApp.clientPrograms[arch].versionRefreshable(),
versionNonRefreshable: Autoupdate.autoupdateVersion || WebApp.clientPrograms[arch].versionNonRefreshable(),
// comment the following line if you use Meteor < 2.0
versionReplaceable: Autoupdate.autoupdateVersion || WebApp.clientPrograms[arch].versionReplaceable()
}
})
const runtimeConfigScript = `__meteor_runtime_config__ = JSON.parse(decodeURIComponent("${encodeURIComponent(JSON.stringify(runtimeConfig))}"))`
const runtimeConfigHash = crypto.createHash('sha256').update(runtimeConfigScript).digest('base64')
const helpmentOptions = {
contentSecurityPolicy: {
blockAllMixedContent: true,
directives: {
defaultSrc: [self],
scriptSrc: [
self,
// Remove / comment out unsafeEval if you do not use dynamic imports
// to tighten security. However, if you use dynamic imports this line
// must be kept in order to make them work.
unsafeEval,
`'sha256-${runtimeConfigHash}'`
],
childSrc: [self],
// If you have external apps, that should be allowed as sources for
// connections or images, your should add them here
// Call helmetOptions() without args if you have no external sources
// Note, that this is just an example and you may configure this to your needs
connectSrc: connectSrc.concat(allowedOrigins),
fontSrc: [self, data],
formAction: [self],
frameAncestors: [self],
frameSrc: ['*'],
// This is an example to show, that we can define to show images only
// from our self, browser data/blob and a defined set of hosts.
// Configure to your needs.
imgSrc: [self, data, 'blob:'].concat(allowedOrigins),
manifestSrc: [self],
mediaSrc: [self],
objectSrc: [self],
// these are just examples, configure to your needs, see
// https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Content-Security-Policy/sandbox
sandbox: [
// allow-downloads-without-user-activation // experimental
'allow-forms',
'allow-modals',
// 'allow-orientation-lock',
// 'allow-pointer-lock',
// 'allow-popups',
// 'allow-popups-to-escape-sandbox',
// 'allow-presentation',
'allow-same-origin',
'allow-scripts',
// 'allow-storage-access-by-user-activation ', // experimental
// 'allow-top-navigation',
// 'allow-top-navigation-by-user-activation'
],
styleSrc: [self, unsafeInline],
workerSrc: [self, 'blob:']
}
},
// see the helmet documentation to get a better understanding of
// the following configurations and settings
strictTransportSecurity: {
maxAge: 15552000,
includeSubDomains: true,
preload: false
},
referrerPolicy: {
policy: 'no-referrer'
},
expectCt: {
enforce: true,
maxAge: 604800
},
frameguard: {
action: 'sameorigin'
},
dnsPrefetchControl: {
allow: false
},
permittedCrossDomainPolicies: {
permittedPolicies: 'none'
}
}
// We assume, that we are working on a localhost when there is no https
// connection available.
// Run your project with --production flag to simulate script-src hashing
if (!usesHttps && Meteor.isDevelopment) {
delete helpmentOptions.contentSecurityPolicy.blockAllMixedContent;
helpmentOptions.contentSecurityPolicy.directives.scriptSrc = [
self,
unsafeEval,
unsafeInline,
];
}
// finally pass the options to helmet to make them apply
helmet(helpmentOptions)
```
### X-Frame-Options
> Note: The X-Frame Options header is configured using Helmet's default header configuration.
From MDN, the X-Frame-Options HTTP response header can be used to indicate whether or not a browser should be allowed to render a page in a ``, `