I really like MongoDB\'s automatically generated ids. They are really useful.
However, is it save to use them publicly?
Let\'s say there is a posts collecti
The ObjectID documentation states that the automatically generated IDs include a 3-byte machine ID (presumably a hash of the MAC address). It's not inconceivable that someone could figure out things about your internal network by comparing those three bytes in various ids, but unless you're working for the Pentagon that doesn't seem worth worrying about (you're much more likely to be vulnerable to something more boring like a misconfigured Apache).
Other than that, Epcylon's right; there's nothing inherently insecure about exposing ids through URLs. Whether it's ugly is another matter, of course. You can base64 them to make them shorter (been thinking about this myself), but then there's the weird fact that they're all about half the same.
I have no experience with MongoDB in a production environment, so don't take my answer as the truth, but I can't imagine why it shouldn't be safe.
Compare to a Auto-ID type column in a RDBMS. You expose those to the outside all the time, I don't know of any reason for not doing the same with MongoDB ids.
As always, the security should be in validating your input and not letting anyone near your database without proper protection. Do it properly and it shouldn't matter if they know how to pick a particular object in your database, as they still can't do anything with it.
Starting with version MongoDB 3.4, MongoDB no longer contains a machine identifier like the marked answer. But the id still contains a timestamp
that can be used to find out when the data was created. If you don't want the ID to store any information related to data that people can use to map data, you can follow what I do.
I have used MongoDB at the production stage. On my system, generally, all MongoDB documents have 2 unique IDs.
First: Default MongoDB _id
(Generated automatically by MongoDB)
_id
(Default)I use this ID to query in the internal system. Examples such as those used for relations between collection
, caching, etc. Querying with ObjectId
is much faster than you querying other data types such as string
.
See https://stackoverflow.com/a/27897720/10861398
Second: Public ID (Generated by your application)
pubId
- You can replace it with anything. (still make sure to use the relevant and concise name)UUID
)A-Za-z0-9_-
)I use this as the main ID when the data is exported out and consumed by the client application. Examples such as data entry and exit between APIs
, integration with third-party applications, as id elements for rows in tables in client applications, etc.
If id gives link to "unlisted" content that requires only link - it's privacy issue.
If id gives a link to content that is under user login - not a problem.
No matter if it's MongoDb, SQL or any other id. Id is the key to data. If this key is only thing you need to view content that you should not - that's an issue. For such situation - generate unguessable id.
I thought mongodb _id was based on a datestamp, and sever address and other things you might prefer to keep private.
If you're worried it might be worth encrypting mongoids and using the result as a client-side identifier (and then un-encrypting when requests come back in).
If the encryption key is partially based on some unique attribute of the user or session in question, that makes it difficult for users to access content when they shouldn't.
Obviously still important to validate the user by other means!
let say we have order 1 by user a order 2 by user b
it's safe to expose the docid for order 1 or 2 for both users ( your code should expose 1 for a and 2 for b though )
but what important is validating
if the user is allowed to read it only and you have like /api/modifyorder/:docid you must validate that the user is an admin
if user B have order 1 ( user A ) docid you need to check the order user id before serving the respond ( and log a security log if he try to do that )