false positives

I have an encryption question: Has anyone tried to create a strong encryption algorithm that cannot be broken by brute force? If the algorithm relied on a dictionary, for instance, then the wrong key could still return something that resembles english. It would be even better if for any arbitrary text, a KEY could be generated which maps the encrypted real message to the arbitrary text. With this system, even if your true key could be figured out (maybe you wrote it on a napkin!), you could easely deny the accuracy of the decryption. A decrypted message would only be useful for information, but never for evidence. This is more of an idea than a question. What do you think? In liberty, Peter Baumbach baumbach@atmel.com