Almost all info related to padding oracle attacks on AES are described in the context of client/server scenarios but does it apply to local encrypted files as well?
For example, a file P is encrypted using AES-256 and CBC padding.
The random IV and C are then written to a new file.
If the attacker got hold of that file, could he then use whatever language he's using to try and decrypt that file and get exceptions that tell him whether he's dealing with padding errors or decryption errors?
If so, how could one protect against this? Wouldn't Encrypt-then-Mac be useless in this case?
Since the attacker can choose to ignore the MAC or am I completely missing something here?