<p>Image encryption plays a significant role in the field of Network Information Security. However, as data volumes surge, the existing encryption schemes have struggled to meet the practical production demands in terms of encryption efficiency and storage burden. To address this issue, this paper designs a chaotic multi-image encryption method by using the spatial permutation and BP neural network compression, which is capable of efficiently encrypting and compressing multiple grayscale and color images of any size simultaneously. Firstly, the spatial permutation scheme is proposed to perform spatial position moving operations on the pixels of the original image cuboid to achieve pixel redistribution. Secondly, the chaotic sequences are utilized to diffuse the scrambled cuboid to obtain the encrypted images. Then, in order to achieve compression, we propose an image compression scheme using the BP neural network which uses the hidden layer data as the compression result and does not require additional keys. We combine the encryption scheme with the compression scheme to achieve image encryption and compression. The plain image cuboid is compressed by the BP neural network to obtain the compression image cuboid, and the compression image cuboid is spatially scrambled and diffused. Finally, experimental results and exhaustive performance analysis demonstrate that our scheme not only has efficient image compression capability but also provides very high security, and can encrypt and compress multiple grayscale and color images of arbitrary sizes with higher robustness and reconstruction quality at the same time.</p>