Is your threat level high enough to use a book cipher?
Traditionally book ciphers work by replacing words in the text of a message with the location of words from the book being used. This can have problems; if a word appears in the text but not in the book, it cannot be encoded.
The main strength of a book cipher is the key. The sender and receiver of encoded messages can agree to use any book or other publication available to both of them as the key to their cipher. The drawback to a book cipher is that both parties have to possess an identical copy of the key. The book must not be of the sort that would look out of place in the possession of those using it and it must be of a type likely to contain any words required.
Messages are transmitted in groups of 3 numbers – 08.23.07.
These are the coordinates that point to a specific word in the book.
08.23.07 gives the location of the word as page 8, line 23, and the 7th word in that line.
You can upgrade this system one more step to transmit a 4 number code to show a single letter. Then 08.23.07.03 gives the location of the letter as page 8, line 23, word 7, letter number 03.
This would make a very long message, that would take some time to construct. BUT, less than the time you would spend as a guest of the government if asked to assist them with their enquiries. In many situations this is the only method that will be available to you, if, as mentioned above, the full word needed to send is not in the book.
If you choose this route then you will obviously not use the same letter code twice in the message. If you are paranoid, as you should be using a system like this, then you should also change the book you use on a regular basis. That will really throw anyone interested a curve – or two, or three.
And then you need a way to send/receive the codes. There are a few ways to do this, including Twitter.