Many of us are aware of that, but are also aware that cyptographic methods are sufficiently well-known world-wide that sigint will be just about useless without humint to back it up.
Also, one point which many people miss is that there exist pencil-and-paper variants on the one-time pad which are more secure than any computer-based encryption (since pencils don't emit decodable RF signatures of what their users are doing).
Although steganography is much more convenient when using a computer than when using pencil and paper, even pencil and paper methods can, when combined with a one-time pad, make it very difficult for anyone listening in to even determine that there is any hidden communication going on and impossible for them to determine what is actually being said.
To clarify what type of system I'm talking about, let me give a (very slightly simplified) example. Before agent Max goes out into the field, Chief gives him one of the two copies of a "code-book" in existence; Chief keeps the other one. Both Max and the Chief keep their codebooks in locked sealed cases whenever they are not in use.
Although the contents are randomly-generated, each page of the code-book has the same format: a list of ten randomly-selected words at the top, preceded by digits 0-9, and then a larger list underneath. The first 26 words in the larger list have the letters A-Z written before them; the later words have a blank in which a single letter may be written in.
When the Chief wishes to send a message to Max, he must first compose the real message and then compose a 'cover' message. Composing the real message is probably the easier part; the only difficulties lie in making it as concise as possible and yet unambiguous. Word spaces and punctuation are elided except where ambiguity would result in which case the letter "X" is inserted. If the large word list on the page has 275 unique words, messages up to 250 letters long may be accommodated, though composing a plausible 'cover message' for something that long may be difficult.
If there were no need for a cover message, Chief would start out by writing out the word corresponding to the first letter of his real message, crossing that word of his list, and writing that letter in the blank next to the 27th word on the list. He would then write out the word corresponding to the next letter of his real message, cross that word off his list, and write that letter in the blank next to the 28th word, etc. The resulting "encrypted" message would be a bunch of random nonsense words which would be completely indecipherable without the a copy of the list that was used in its encryption.
If a cover message is needed, it must be carefully crafted so as to use all the necessary words in order, with whatever arbitrary stuff the Chief feels like putting in between them, with one important caveat. At any point when writing out the message, 26 words will be 'active'; the extra nonsense thrown in between words of the encrypted message must not use any of the active words or the resulting message will be garbled when Max tries to decode it.
Once Max receives a message, he decodes it using the appropriate page of his code book: he checks the words of the message one by one against the first 26 words in the code book list until he finds a match. He then crosses off that word and writes the letter that was in front of it in the blank before the 27th word. He then continues checking words of the incoming message until he finds another match; he writes the letter in front of that word in the blank before the 28th and continues in this fashion until he has gone through the entire message. Once he has decoded the message, Max then takes out his zippo and destroys the page he used to decrypt it, since he will never need it again.
If the word lists are truly randomly generated, it will be literally impossible for anyone intercepting the messages to decode them without a copy of the lists. Since every page has a different random list and is only used for a single message, deriving the contents of a page based upon knowlege of the plaintext and ciphertext will be useless. Additionally, since each page of the codebook is destroyed after use, capturing the code book would not allow someone to decipher previously-sent messages.
The one possible weakness in this system would occur if someone captured Max's codebook without Chief knowing about it. This possibility may be protected against, however, if Max and the chief have agreed on a set of number sequences for each day of the week (assuming that any two consecutive messages will be sent on different days of the week). Before the Chief reads max his encrypted message, he speaks a sentence containing the word from the top list corresponding to the first agreed-upon number (and no other word from that list); Max responds with a sentence containing the word for the second number, Chief replies with the third, etc. If Max (or an imposter claiming to be Max) fails this handshake sequence, Chief will know not to entrust him with any more communications using that codebook.
If this protocol is followed, then unless Kaos can copy the codebook without Max or the Chief being aware of it, the most Kaos would ever be able to gain would be a single intercepted message if they raided Max while he was receiving it. Although they would know the number sequence for the current day of the week, they would have no way of knowing the number sequence for any other day and thus no way of impersonating Max. Consequently, no future messages would be sent.
As should be obvious, the above communications method would probably be annoyingly clumsy to actually use; computers are much more convenient. On the other hand, provided Max was given enough code-book pages to handle all the messages he'd ever have to receive, the method would be completely unbreakable unless Max or the Chief allowed someone to copy their codebooks.