So I'm making an app where I want the users to be able add, edit and rate content, but I do not want to force them to register. Instead I was planning on just using their device id or device token to identify them. I'm planning on making both an iPhone and Android version, so I'm looking for a general solution, but the iPhone version has higher priority, so an iPhone specific solution would also be welcome.
The problem is that I don't want just anyone to be able to use my web service by sending a phony device id or someone else's device id.
How would the client prove to the server that it is providing the correct device id?
In theory, you cannot. A device ID is not particularly secret, and in most cases, it can be easily spoofed. As for Android, there's no reliable device ID on that OS at all - see the gory details here: Is there a unique Android device ID?
All you can rely upon is security by obscurity - hoping that no one will be determined enough to reverse-engineer the code and analyse the authentication protocol. And not disclosing the code is not an option - you are distributing the app after all.
That said, one not-particularly-secure auth method would be - send the device ID and a hash of device ID concatenated with a secret, hard-coded in the client code string (the shared secret). The service would contain a copy of the secret, recalculate the hash (using the device ID provided) and match the hashes. Not breakable by protocol analysis, only by digging in the code for the secret. Vulnerable to replay attacks though. Feel free to obfuscate the secret in the code - e. g. combine it together from parts stored in separate places right before use.
For a stronger solution, authenticate users, not devices. This is up to your customers, and depends on the nature of the business.