gwyn@brl-smoke.ARPA (Doug Gwyn ) (02/28/88)
In article <1857@phoenix.Princeton.EDU> amlovell@phoenix.Princeton.EDU (Anthony M Lovell) writes: >As long as your method of devising a key from the plaintext is arbitrary >(and remains unknown to the cracker) , he will not get his foot in the >door. What if the n+5th message is the key (again adulterated in some >form) for the nth message? Any scheme like this will be impregnable >until guessed, and its patterns are certainly unlike those typically >searched for. The arbitrary system can be changed in encrypted >transmissions (with the acknowledged risk that it will not help IF >the cipher is already compromised). This denies the "enemy" a large >body of ciphertext to examine for these weak patterns. First, the system originally described was indeed a simple auto-key system that any competent cryptanalyst could quickly break. Second, in your changed scenario, the whole system collapses if any of the message traffic is not received. In practice, for important cryptosystems (as opposed to toys amateurs play with), you have to count on the general system being "somehow" known to the opposition. You also need to include sufficient indicators in the transmitted text to allow the receiving station to select the right decryption key without fumbling around. Of course these indicators will also be visible to the opposition's cryptanalysts. > Not by any means the most secure or practical system (more secure than >practical in my mind), but I would put my money on the cleartext >remaining undiscovered for a LONG LONG time. If there were any reason for competent cryptanalysts to concentrate effort on your system, you'd lose your money (not to mention your secrets).