Dresden 2003 – scientific programme
Parts | Days | Selection | Search | Downloads | Help
DY: Dynamik und Statistische Physik
DY 42: Synchronisation I
DY 42.3: Talk
Thursday, March 27, 2003, 11:30–11:45, G\"OR/226
Synchronization of Random Walks and Chaotic Maps: Application to Public Channel Cryptography — •Wolfgang Kinzel1, Ido Kanter2, Rachel Mislovaty2, and Andreas Ruttor1 — 1Theoretical Physics, Am Hubland, 97074 Wurzburg — 2Theoretical Physics, Bar Ilan University, Israel
Two artificial neural networks (parity machines or random walks) can be synchronized by mutual learning. After synchronization, their common synaptic weights can be used as a secret key generated via a public channel. Recently, Shamir et al suggested several attacks on neural cryptography. We show that the probability of a successful attack can be made exponentially small with the number of bits per weight. The security of neural cryptography can be increased by two methods: First, by combining neural networks with chaotic maps which are synchronized by the networks, and second, by using a feedback input.
References and preprints can be found in
http://theorie.physik.uni-wuerzburg.de/TP3/publi.html