A New Security on Neural Cryptography with Queries

We can generate a secret key using neural cryptography, which is based on synchronization of Tree Parity Machines (TPMs) by mutual learning. In the proposed TPMs random inputs are replaced with queries which are considered. The queries depend on the current state of A and B TPMs. Then, TPMs hidden layer of each output vectors are compared. That is, the output vectors of hidden unit using Hebbian learning rule, left-dynamic hidden unit using Random walk learning rule and right-dynamic hidden unit using Anti-Hebbian learning rule are compared. Among the compared values, one of the best values is received by the output layer. The queries fix the security against majority flipping and geometric attacks are shown in this paper. The new parameter H can accomplish a higher level of security for the neural key-exchange protocol without altering the average synchronization time.