Security By Obscurity — a New Theory
mikejuk writes "Kerckhoffs' Principle suggests that there is no security by obscurity — but perhaps there is. A recent paper by Dusko Pavlovic suggests that security is a game of incomplete information and the more you can do to keep your opponent in the dark, the better. In addition to considering the attacker's computing power limits, he also thinks it's worth considering limits on their logic or programming capabilities (PDF). He recommends obscurity plus a little reactive security in response to an attacker probing the system. In this case, instead of having to protect against every possible attack vector, you can just defend against the attack that has been or is about to be launched."
I hate it when people always seem to take the phrase out of context and apply it to mean any kind of security, like network security or the old Windows/Linux battle. It's a completely different kind of situation, and in the former it's especially true that security by obscurity is a hardener layer. It's also why Linux has managed to stay as (consumer) malware free to day, even though it still has a fair share of its own worms and other security problems.
Which means that the real security is the lock on the door.
But that is also just obscurity in another form. The obscure part is that the attacker doesn't know the combination to the lock, or doesn't know how the tumblers specifically are keyed. Otherwise a key could be made up.
All security is obscurity, just different levels of it. In some schemes the obscure value is shared (hidden directory on the server that isn't crawled but can none the less be accessed by a direct link). Some obscure values aren't (public key encryption).
The hiding the key under the rock is analogous to using a weak form of obscurity to hide a strong one. Which in this case is no better than the obscurity of not letting anyone know that the door lock doesn't actually work anyway.