The trend prediction of software vulnerability can provide valuable threat intelligence in security event prevention. It is a challenging task for highly accurate prediction. To address this problem, a novel prediction method, STL-EEMD-ARIMA, is proposed, by incorporating Seasonal and Trend Decomposition using Loess (STL), Ensemble Empirical Mode Decomposition (EEMD), and Autoregressive Integrated Moving Average (ARIMA). Firstly, the trend and random fluctuation components are extracted from the vulnerability trend samples by using the STL method, respectively. Secondly, we use EEMD to decompose the trend and the residual to achieve Intrinsic Mode Functions (IMFs). Then, the IMFs are classified into high- and low-frequency components based on the data characteristics. Finally, the ultimate prediction result is obtained by ARIMA. The experimental results of the proposed method illustrate that the absolute error rate decreases by an average of 4.22% compared with the naive EEMD-ARIMA method.