To fulfill the requirements of data security in environments with nonequivalent resources, a high capacity data hiding scheme in encrypted image based on compressive sensing (CS) is proposed by fully utilizing the adaptability of CS to nonequivalent resources. The original image is divided into two parts: one part is encrypted with traditional stream cipher; the other part is turned to the prediction error and then encrypted based on CS to vacate room simultaneously. The collected non-image data is firstly encrypted with simple stream cipher. For data security management, the encrypted non-image data is then embedded into the encrypted image, and the scrambling operation is used to further improve security. Finally, the original image and non-image data can be separably recovered and extracted according to the request from the valid users with different access rights. Experimental results demonstrate that the proposed scheme outperforms other data hiding methods based on CS, and is more suitable for nonequivalent resources. encryption together through matrix multiplication [Rachlin and Baron (2008)]. Compared with the stream cipher encryption, the encrypted data based on CS can reduce bandwidth resources effectively. Due to this feature of CS, research results based on CS often serve multimedia data compression and representation [Wu, Yu, Yuan et al. (2016)]. Meanwhile, CS has a low computation cost in the sensing part while the computation of recovery is rather complex at the receiving end. Information hiding plays an important role in protecting various information from being destroyed [Cao, Zhou, Sun et al. (2018)]. Within the current application of CS in reversible data hiding, there are two main types: the first one usually embeds data in the samples of DCT/DWT and then uses CS to compress [Xiao and Chen (2014)]. The other one embeds data in the measured value [Cao, Du, Wei et al. (2016); Li, Xiao and Zhang (2016); Pan, Li, Yang et al. (2015)]. However, both of the two schemes have some defects. The first one is more suitable for digital watermarking rather than data hiding, which focuses on the protection of the copyright about the carrier and the robustness of the watermark. But, data hiding puts more emphasis on the capacity and security of the embedded information itself. The scheme proposed by Pan et al. [Pan, Li, Yang et al. (2015)] is a watermarking scheme for plain image only which does not provide security for the cover. The scheme proposed by Cao et al. [Cao, Du, Wei et al. (2016)] is not suitable for nonequivalent resources, although this scheme has a brilliant performance on recovery. The reason is that sparse representation to vacate room for data embedding in the preprocessing operation is very complicated. The scheme proposed by Li et al. [Li, Xiao and Zhang (2016)] is a smart data hiding scheme based on block compressive