This is the contradiction built into cryptanalysis. It is useless to "break" a cipher system if those who use it know you have 'broken' it. But cryptanalysis requires the cooperation of many, and is therefore an intrinsically insecure process.
In September 1941 a Japanese destroyer came into the harbor at San Francisco. As was traditionally when foreign warships visited, a naval inspection party went aboard. The ship's security was sloppy, and a naval officer found himself all alone passing the radio room, which was unoccupied and its door open. And there on a table lay a copy of the Japanese Naval Code Book. He picked it up and returned with visions of greatness dancing in his eyes. He's probably still stuck somewhere north of the arctic circle guarding an igloo or something. It so happened that U.S. cryptanalysts had broken that code -- which of course the Japanese immediately stopped using when they discovered it had been compromised. All that work to do over again.
Carrol