Abstract:Grey self-organizing map (GSOM) model is proposed and applied in the detection of intrusion. Through the improvement of the weight adjustment using the GRC (grey relational coefficient), the training results of SOM get better. In the detection of deny of service (DOS) attacks, this model can consider the relativity of the data set of DOS attacks. Finally, the experiments on the DOS data set confirm their validities and feasibilities over this GSOM model. This work has been supported by the Tianjin Key Subject Fund, and the Higher Education Science and Technology Development Fund of Tianjin Municipal Education Commission (No. 2006BA19).