ENet packets security

[amaury-s]

Hi,
I'm working on a little multiplayer game (using Godot and therefore ENet) and I tried to figure out what makes ENet secure for an host. I searched in the code but couldn't find the answer.
If I understand correctly, ENet does not handle requests but instead Packets? If this is correct, how does ENet handle packets that are not made by a legitimate client (using my game) but instead a malicious actor that try to take over the host?

Hope it makes sense and sorry if it is not the place for this question.

[Faless]

@amaury-s for questions about godot games that happen to use ENet, please ask your questions in the Godot discord, Godot forums, or even the Godot developer chat if you feel it's very technical and can't get answers from the forums or discord,

Godot also uses a slightly modified version of ENet that adds some features, so the upstream might not be able to help you much.

That said, ENet has no way to know if a packet comes from your game, or a different software that happen to use the same packet format.

There's no easy way to make sure only valid clients can connect, you can ensure that valid clients can only connect to valid servers by using DTLS, if you don't distribute the server itself.

Adding client certificate verification on the server doesn't help per se, since the moment you embed a client certificate in the client, anyone who has access to the client will be able to extract it, and use it to authenticate with the server.

You can investigate some fancy setup with external infrastructure that generates certificates / authentication codes after requiring a user login, but more or less the same problem would apply (i.e. malicious users will be able to extract the certificate, and use it in a malicious client). The only plus is that you'll be able to know which user certificate / code is being used, and potentially ban the given user once / if you realize they are doing that.

You can also use an authentication function and DRMs from your distribution platform (e.g. the Steam DRM), but again, it can (and probably will) be broken at some point.

take over the host?

I am not sure what you mean by "take over the host", but in general, unless you code your server to allow them to "take over the host" (e.g. executing shell command from an RPC without proper validation), clients will not be able to interact with the host machine directly, and will only be able to call what you expose on the server.

[amaury-s]

Hi @Faless, thank you for your answer. I will ask in the specific forum you sent!

[amaury-s]

Also concerning ENet, does it prevent memory/buffer overflows somehow?