@i_nardi There is an obvious fix to that. Create a post encryption stage that derandomize the output. Something like Huffman encoding, but in reverse. Make it look like English plain text, or Chinese plain text. Of course, it becomes longer...