Encoding cash, the details

 One card send cash to the other. The first card has a map of the second cards personal iD, it never needed the full secret key, just the map.

The map is approximately binomial and centered, the second card had trained on that key.  The cash to be sent is text, and will e encrypted with the map of the destination. Ifthe message is encoded, then it too9 is approximately centered binomial.  I can encode the message with a map that exactly scales to the key map, then multiply them and send the diagonal. Keep the ratio of key histogram matched to the message histogram.  

My message is finite, but can be very long. I make the minimum Huffman tree to separate the bytes. Like a pit boss, i add my market grease, but send the market risk to the destination.  That is the difference sequence, the differential map that lets the decoder separate key from message.. 

Explain this.  A message might be a thousand bytes long, and the encrypting key only 256.  If if I take the message Huffman encode, I  get an unfair coin toss that does not match the key binomial.  I want to warp the encoding ree to match the key binomial and there should be one optimum solution.  For all those node alterations, I need to tell the destination because it hast to decode to four times the accuracy.   The encoder then need only send the two bits extra of fraction to sub select the proper message code. The message map is simpler, It has an offset and slop relative to the key map. Both of them optimum approximations of Gauss.

The attacker has no idea what the probability of a byte coded key value, it cannot reconstruct the binomials.  But the message is not making maximum use of bandwidth,  two of he bis are not compressed really, and a lot of empty space was left in the encoding tree as it mapped to the key binomial.  I can expand the resolution of the message encoder as needed to drop congestion probability.