50g: an interesting RAND anomaly

03192018, 06:31 PM
Post: #17




RE: 50g: an interesting RAND anomaly
One must be careful with randomly programmed random number generators. When I was in grad school, one of my professor came running down the hall, flagged some of us down and showed his newest result. He and another prof ha proved that the average length of all mappings from the integers 0 to N1 mod N was Sqrt(N).
A practical application came about 30 years later. A group designing some type of password stuff came to me with their modification of DES and asked me to determine how many different keys would result if the put in 10 digit passwords. I also got to do something I had wanted to do; that was, write a Fortran code that simulated DES and kept track of where each bit went. I could figure out which bits depended on which plaintext and which key bits. I told the group that if they just fiddled around with 10 digit passwords and put them through something that wasn't designed to handle 10 digit objects, they would get about 100,000 different outputs. Needless to say they wanted a simulation (which I had) so I ran a few million test cases. My simulation code was nearly as fast as many streamlined DES implementations. The idea was to feed the plaintext back into system as cypher text and look for a repeat. The average cycle length was 10^5. I do not know if they changed anything based on this result. There was the theory and simulation (actually, a bitforbit emulation) of their system. These two answers agreed. 

« Next Oldest  Next Newest »

User(s) browsing this thread: 1 Guest(s)