I am not sure if it is pre-mature to ask such a question. The question is largely inspired by pondering over the question and discussions in the following link (In his answer, @Dad also raises similiar question there):
Can we still provide confidentiality when cryptography is outlawed?
Question: Are there any theoretical models of undetectable encryption system or any elementary work done on the same?
Most modern encryption systems are detected because it is easier to know that they are encrypted. For example:
/*a basic working of an encryption function*/ char* a = encrypt("Hello World"); printf("%s", a);
We get something like, a = #$%HGFGYTU@724. Most security experts who intercept the exchange of 'a' will realise that it is an encryption and proceed with whatever further action. But on the contrary if 'a' was encrypted something like "I am fine" then, at first place, it would be difficult for the eavesdropper to even detect a presence of encryption. Thus, 'making sense' of encrypted text probably may be important.
By using the phrase "undetectable encryption system" I am explicitly ruling out mechanisms like steganography, for they often work on the probabilistic models of information being located or decoded.
I am talking of a deterministic model of encryption where if C eavesdrop on a message from A to B then C receives an interpretation, which though completely sensible to him/her, is not the same as the the interpretation shared by A & B.