From what I understood, one of the advantages of quantum communication is that one can (mathematically/physically) prove that the quantum communication/message has not been intercepted/tampered with. (I believe this proof relates to the no-cloning theorem).
However, I would like to ask: Is it possible (in theory) for Alice and Bob to communicate using quantum communication techniques without anyone being able to detect that they communicate? And if yes, is there a cost to such a level of security?
I assume Alice and Bob are able to meet physically and undetected to set up some key/pairing/initial encryption. (I am not sure whether this is required, yet I assume so.)
Regardless of whether Alice and Bob share some amount of entangled states, there is no way for them to communicate without having Alice physically sending something to Bob. "Something" means here classical information, or quantum states alike. Think of them sharing entanglement as their states being correlated in some (more or less "special") way. Them sharing correlated objects doesn't mean in any way that such correlation can be used to communicate information.
So, whether someone might or might not be able to detect the "package" Alice is sending to Bob has nothing to do with no-cloning, or QM at large really. There is no fundamental way to prevent anyone from even in principle intercepting the states sent from one side to the other.
You don't need quantum in order to communicate while evading detection. There's nearly a century of research on what is known as LPI/LPD (Low-probability of intercept/low-probability of detection) communication. The basic idea is you hide your message signal in a way that looks like background noise to anyone trying to intercept the message.
If you already have a pre-shared key, this becomes even easier. For example you can spread your signal over a very broad bandwidth such that you only recover the signal if you know how to filter/transform the broadband signal (known via your shared secret key) and otherwise any measurement of the signal just looks like background EM noise. Likewise you could randomly hop between different channels in a way that is only predictable if you know the secret info of how you channel hop. For a short intro (and other examples), see here.