I often warn about the dangers of blindly trusting user provided data. There are times, though, when it’s easy to forget where your data comes from. Sir Charles Watson brought up a good example of this when he found a security issue revolving around an insecure onCreateUser hook.

Imagine that somewhere in your application you’ve defined an onCreateUser callback like the one below:

Accounts.onCreateUser(function(options, user) {
  user.isAdmin = !!options.isAdmin;
  return user;
});

The isAdmin flag is set in a method used by administrators to create new admin users:

Meteor.methods({
  createAdminUser: function(email) {
    check(username, String);

    var user = Meteor.users.findOne(this.userId);
    if (user && user.isAdmin) {

      var newUserId = Accounts.createUser({
        email: email,
        isAdmin: true
      });

      Accounts.sendEnrollmentEmail(newUserId);
    }
  }
});

This looks secure enough. We’re asserting that the username is a String, and that the currently logged in user has the appropriate authorization to create a new admin user. No one could tamper with this method and fraudulently create an admin account.

Unfortunately, we’re forgetting that createUser can also be called from the client. Additionally, any client can provide their own options argument. An in-the-know user could create an admin account by running this code in their browser console:

Accounts.createUser({
  username: "loladmin",
  password: "loladmin",
  isAdmin: true
});

After running the above line, the user would be logged in and granted full administrator permissions.


There are two obvious fixes to this security issue. The first, and by far the most straight-forward is to simply disable user account creation on the client using forbidClientAccountCreation. This may not be the best solution, though, as it would prevent users from signing up with your application. Only administrators would be able to create users through server-side methods.

A better solution may be to remove the onCreateUser callback and move the admin creation functionality into an entirely separate method:

Meteor.methods({
  createAdminUser: function(email) {
    check(email, String);

    var user = Meteor.users.findOne(this.userId);
    if (user && user.isAdmin) {

      var newUserId = Accounts.createUser({
        email: email
      });

      Meteor.call(“makeUserAdmin”, newUserId);
      Accounts.sendEnrollmentEmail(newUserId);
    }
  },
  makeUserAdmin: function(userId) {
    check(userId, String);
    var user = Meteor.users.findOne(this.userId);
    if (user && user.isAdmin) {
      Meteor.users.update(userId, {
        $set: {
          isAdmin: true
        }
      });
    }
  }
});

Using this solution, users can still register for your application using createUser on the client, and admins can still create new admin users using the createAdminUser method.


This issue is indicative of a broader issue with developing Meteor. Developers often forget that most of the code they’re writing is isomorphic. Most of the code you write can (and will be) run on both the server and the client.

Sometimes it’s easy to lose track of where your code can be executed. Is the file you’re editing buried deep within a server folder? Is the code you’re writing wrapped in a Meteor.isServer guard? Is the code you’re writing in any way triggerable by a client-side action? These questions can be cognitive burdens, but they’re important to keep in mind at all times when developing secure Meteor applications.