Showing posts from June, 2024

Neural PRNG

Neural Network in PRNG The greatest share of PRNG vulnerability is created by constant parameters: constant (even if considered crypto-resistant) generation algorithm, constant length of generated numbers, constant (on some interval of generated numbers) master key. Using neural networks with feedback in PRNG leads to inflation of crypto-resistance. With the help of neural networks it is easy to make ALL generator parameters variable. At every step. The complication of cryptanalysis is especially effective with preliminary distortion of encrypted data. For example, when using several layers of encryption. Of course, we are not talking about stream encryption. Let's write an example of a neural network in Java from one neuron, emulating a network of 6 neurons. Then we'll teach it to rebuild the structure and change parameters. But for now we won't study learning. Android JAVA