This paper presents a method for distributed attacks detection in corporate computer networks. The method is designed according to the modern requirements of attack detection and can be used to reveal existing and modified attacks. A mathematical model of the system is developed using hybrid neural networks. Hybrid networks are presented as an integration of two artificial networks – a multilayer neural network and a self-organizing map or Kohonen network. The model can be applied in corporate networks for increasing efficiency of security management. Corporate networks formed the environment of network economy and disseminate critical managerial, business and financial information, which needs to be protected from dynamically changing computer attacks. The developed methodology was applied to analysis of the Armenian commercial banking system.