In view of the problems that the current data security monitoring methods do not consider the dimensionality reduction of dynamic shared data, resulting in long monitoring time, low monitoring accuracy and poor selection effect of data dimensions and feature attributes, a dynamic shared data security monitoring method based on CP-ABE is pro-posed. The stationary distribution of Markov chain and the encryption characteristics of CP-ABE algorithm are analyzed. The noise in dynamic shared data is eliminated according to low rank and sparse decomposition. The linear mapping matrix is constructed to com-plete the dimensionality reduction of dynamic shared data. The artificial fish swarm algo-rithm is used to optimize the secondary feature selection of dynamic shared data. Through the density peak clustering algorithm, the dynamic shared data is divided into clear data points and fuzzy data points, the candidate class labels of fuzzy data points are obtained, and the hybrid leapfrog algorithm is used to determine the final class label to realize the security monitoring of dynamic shared data. The experimental results show that the data dimension and feature attribute selection effect of the proposed method is good, which can effectively reduce the monitoring time of dynamic shared data and improve the monitoring accuracy of dynamic shared data.