The proliferation of high-definition cameras in smartphones and other portable devices has led to an explosive growth in personal image and video data. Cloud storage has become indispensable for managing these vast collections, enabling convenient access and sharing across devices. However, such images often contain sensitive personal information, making them targets for external attacks or internal breaches, as illustrated in Fig. 1(a). A common countermeasure is to encrypt images before uploading, ensuring that only ciphertext is exposed to adversaries or administrators (Fig. 1(b)). While traditional encryption secures content, it completely obscures visual information, forcing users to decrypt entire libraries for browsing—severely undermining usability [1], [2], [3], [4], [5], [6], [7], [8]. This tension highlights the critical need for encryption schemes that preserve both privacy and practical image usability.

As compact visual representations, thumbnails enable efficient cloud-based image retrieval by minimizing bandwidth and computational overhead. Critically, cryptographic schemes that preserve thumbnail usability while restricting adversaries to only these limited previews can significantly enhance security over plaintext storage, achieving an optimal balance between privacy and accessibility (Fig. 1(c)).

Introduced in 2015, the TPE concept was first presented by Wright et al. [9]. Their scheme used pixel scrambling while preserving pixel sum averages to maintain thumbnails precisely, though it leaked original pixel values. The subsequent research landscape has evolved into two main branches: ideal TPE and approximate TPE. Ideal TPE, building on Wright’s work, was advanced by Tajik et al. [10], who constructed the first theoretically secure scheme using a substitution-permutation paradigm. Enhancements followed, including Zhao et al.’s TPE2 using pixel triplets [11], and Zhang et al.’s integration of chaotic systems for improved efficiency [12]. Recent innovations include genetic algorithm-based [13], GAN-based [14], [15], and various chaotic system-based schemes [16], [17], [18], [19].

In approximate TPE, Marohn et al. proposed efficient schemes like TPE-LSB and DRPE [20]. Zhang et al. later developed a high-fidelity scheme (HF-TPE) using Sum-Preserving Data Embedding (SPDE) [21]. Further works integrated reversible data hiding for perfect recovery [22], enabled encrypted image comparison [23], combined compression [24], employed MSB prediction for noise-free images [25], utilized Markov models [26], and integrated DNA encoding with difference expansion [27].

To ensure that encrypted images reveal only thumbnail information without disclosing any additional sensitive data, while simultaneously enabling perfect reconstruction of original images through the decryption process, this paper proposes a novel approximate TPE scheme. The proposed framework integrates reversible data hiding techniques with image encryption based on the Lorenz hyperchaotic system and RSA algorithm [28]. Recently, the field of Thumbnail-Preserving Encryption (TPE) has witnessed a series of innovative research contributions that further enrich its theoretical framework and application scenarios. The E-TPE scheme [29] proposes a thumbnail-preserving encryption method based on Efficient Rank Mapping, which reduces the computational overhead of encryption by optimizing the pixel value sorting process. The PR3 framework [30] integrates reversible data hiding with privacy enhancement mechanisms, presenting a dual protection architecture that offers both recoverability and usability enhancement. The TPE-DF scheme [31] achieves thumbnail generation in the encrypted domain through Dual-2D Compressive Sensing (Dual-2DCS) fusion, demonstrating high encryption efficiency. The MLVPP scheme [32] introduces the concept of Multilevel Visual Privacy Protection, enabling fine-grained access control through thumbnail preservation and key sharing mechanisms. Specifically, our approach innovatively exploits the edge regions of images for reversible data embedding. By concealing recovery information in these peripheral areas via reversible data hiding, the scheme slightly degrades the visual quality of the encrypted image’s edges, yet guarantees lossless reconstruction of the original content upon authorized decryption. This hybrid design aims to achieve an optimal balance between security guarantees and practical usability in cloud-based image storage environments.

This work makes the following contributions:

1. We propose a novel approximate Thumbnail-Preserving Encryption scheme that integrates the RSA algorithm with the Lorenz hyperchaotic system for enhanced image encryption. This hybrid cryptographic approach ensures that only thumbnail information is revealed in the encrypted domain while maintaining strong security guarantees.

2. We develop a unified framework that combines image encryption with reversible data hiding techniques, enabling complete restoration of processed images to their original pristine state. This dual-functionality architecture provides both confidentiality preservation and perfect reconstructability.

3. Our proposed framework is characterized by leveraging the image’s peripheral regions for reversible data embedding. This design strategically trades off marginal visual quality in the edge areas to guarantee lossless decryption, thereby achieving a novel balance between security and practical usability.

The remainder of this paper is organized as follows. Section 2 outlines the preliminary work and foundational concepts relevant to our scheme. Section 3 provides a detailed description of the proposed encryption methodology. Section 4 elaborates in detail on the decryption scheme corresponding to the encryption scheme. Section 5 presents comprehensive experimental results and performance analysis. Finally, Section 6 concludes the paper with a summary of findings and potential future research directions.