Verifiable and visually meaningful image encryption scheme based on two-stage data fusion in the encrypted domain
Haojie Li, Meng Li · Journal of Electronic Imaging · 2025
Visually meaningful image encryption (VMIE) schemes disguise the secret image as another common image, effectively avoiding hackers’ notice and suspicion. Most existing schemes embed pre-encrypted secret images into the plaintext domain of host images. This may lead to the following security risks: (1) the easy accessibility of embedded data through direct statistical analysis of ciphertext images and (2) visual quality degradation caused by direct embedding images with different distribution features. In addition, existing schemes rarely consider identity verification for the data-sharing party. Motivated by these shortcomings, we develop a verifiable VMIE framework based on data fusion in the encrypted domain. First, an improved chaos system is designed to effectively pre-encrypt the digitalized handwritten signature, secret image, and host image. Then, an embedding strategy of two-stage data fusion in the encrypted domain is proposed to integrate the handwritten signature, secret image, and host image by least significant bit-based data hiding. Finally, the ciphertext image is reconstructed by applying the reverse operation of pre-encryption to the fusion result. Different from general schemes that directly embed noise-like images into the plaintext domain, our data fusion is completed in the encrypted domain. Due to the similarity of images in the ciphertext domain to noise images, the secret and host images after pre-encryption have close distribution and texture features. Thus, it reduces disturbance to the host image caused by secret embedding, thereby improving the visual quality of the ciphertext image and the security of the secret data. Moreover, the introduction of digitalized handwritten signatures has brought a more convenient user experience for identity verification. Experiments show that the proposed scheme can recover the secret image and handwritten signature even after undertaking various attacks. Compared with the state-of-the-art algorithms, the proposed scheme has a higher visual quality of ciphertext images and recovered secret images.