Cooperating attackers in neural cryptography

An attack strategy in neural cryptography was described. The neural cryptosystem, which is based on synchronization of neural networks by mutual learning, was found to be secure under different attack strategies. The 'majority-flipping attacker' does not decay with parameters of model. This attacker's outstanding success is due to its using a group of attackers which cooperate throughout synchronization process. An analytical explanation of this attack is also presented.