Comment by Jarwain | Hacker News Reader

Comment by Jarwain | Hacker News Reader

Parent Full thread

Idk that reeks of security through obscurity to me. Your authorization/permission scheme has got to be fubar'd if you're relying on obscuring IDs to prevent someone from accessing a resource they shouldn't.

I'm sure I'm missing something obvious, but I'm not sure what other threat vectors there are, assuming it's in conjunction with other layers of security like a solid authorization/access control scheme.

I guess I'm not the biggest fan for a few reasons. I'd rather try and design a system such that it's secure even in the case of database leak/compromise or some other form of high transparency. I don't want to encourage a culture of getting used to obscurity and potentially depending on it instead of making sure that transparency can't be abused.

Also, it just feels wasteful. If you have two distinct IDs for a given resource, what are you building your foreign keys against? If you build it against the hidden one, and want to query the foreign table based on user input, you've gotta either do a join or do a query to Get the hidden key just to do another query. It just feels wasteful.

EDIT: apparently RFC 4122 even says not to assume UUIDs are hard to guess and shouldn't be used for security capabilities. So if it shouldn't be depended on for security, why add all this complexity to keep it secure?

The point you may be missing is that the key itself contains information about records in the database that you don’t have access to. There are many famous examples in literature (e.g. the German Tank Problem [0]) of useful attacks on known unique serials to infer hidden information without access. In a database context, the keys for the records you are authorized to see can tell you much about the parts of the database to which you have no access.

Strong randomization of record serials mitigates these attacks.

[0] https://en.wikipedia.org/wiki/German_tank_problem

I thought we were talking about UUIDv7, which is random enough to make this not a problem right?

The idea being to expose uuid instead of the natural index.

It’s been downgraded as people use uuids more.

That said, security through obscurity is an effective layer, particularly for slowing an attack.

Slowing lateral movement is valuable.

Sorry I'm a bit confused.

I'm in agreement that a natural key shouldn't be used as the primary key for a record.

I was responding to a comment about having a hidden "database ID" (which I interpreted as being a serial key?) and a public "Uuid", and questioning the utility of that hidden database ID versus having a public UUIDv7 as the sole primary key, followed by questioning whether the utility of obscuring that primary UUIDv7 is worth the complexity of having to manage multiple artificial keys.

I agree that security through obscurity is a valuable layer in a multi-layered security position.

I guess I just don't think obscuring a Uuid primary key is worth the added complexity in most systems.

I see it like adding a second front door to your house with a separate set of keys. Sure it'd be more secure, but it's an added pain and doesn't help if you don't have a sturdy doorframe, or smash-resistant windows.

UUID is PUBLIC id, use it to look up the bigint numerical id, when necessary.

one should not divulge scale, placement in numerical sequence, etc wtr to integer id, hence ouvlic UUID, which is basically unguessable token

Why have a bigint numerical ID at all?

fast lookup for internal usage. it's significantly faster than a UUID on every DBMS i have ever seen

That's a good reason! I think with UUIDv7, because it's sequential it's indexes are faster than UUIDv4. Still larger than bigints though. I'd like to do a benchmark at some point

The only thing UUIDv7 exposes is its creation time, which isn't tremendously useful or secret information.

In this case obscurity is security: leaked information from the database is directly useful and a secret in itself, not only a stepping stone that might or might not facilitate further exploitation.

Reply on news.ycombinator.com