Abstract: To achieve better results in the individual assessment and the group index construction of the cyber security awareness, a high-accurate method was proposed, combining with a set of different techniques. On the first stage, multi-dimension data acquisition and mixture of regression were introduced for the quantification of the security score. And then, gathering the extracted data with the improved label propagation algorithm, the awareness level was obtained based on the customized supervising data, being independent of tagged data. Furthermore, by using the Grubbs criterion to design the calculation function for group security index, to diminish the abnormal points and abstract the common features, the group index was constructed based on the evaluation results of cyber security awareness. Compared with the selected state-of-the-art counterparts, the proposed design can achieve a better accuracy and objectivity, and be able to demonstrate the general standard for the group security index. The results provide a effective method for the further study of the cyber security awareness.