<p>Existing state-of-the-art secret sharing schemes primarily focus on generating shares for the entire image. However, there is limited research in the literature that specifically focuses on objects or attention-worthy objects (salient objects) present in an image. Performing share generation and reconstruction over the complete image often leads to increased computational, communication, and storage overhead due to redundant background information. To overcome these limitations, this work proposes a visual saliency-based multi-secret sharing (VS-MSS) scheme for secure and efficient cloud storage, focusing exclusively on salient objects as the actual secrets. The proposed framework integrates a tri-stage encoder--decoder network for salient object detection with a lightweight arithmetic modulo-based secret sharing mechanism for efficient share generation and reconstruction. The method utilizes operations such as random permutation, modular arithmetic, and modular multiplicative inverse, achieving an optimal balance between data security and computational efficiency. The scheme follows a perfect <InlineEquation ID="IEq1"> <EquationSource Format="TEX">$(n + 1,n + 1)$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mo stretchy="false">(</mo> <mi>n</mi> <mo>+</mo> <mn>1</mn> <mo>,</mo> <mi>n</mi> <mo>+</mo> <mn>1</mn> <mo stretchy="false">)</mo> </mrow> </math></EquationSource> </InlineEquation> multi-secret sharing configuration, where the first <InlineEquation ID="IEq2"> <EquationSource Format="TEX">$n+1$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi>n</mi> <mo>+</mo> <mn>1</mn> </mrow> </math></EquationSource> </InlineEquation> represents the number of generated shares distributed across independent non-colluding cloud nodes, and the second <InlineEquation ID="IEq3"> <EquationSource Format="TEX">$n+1$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi>n</mi> <mo>+</mo> <mn>1</mn> </mrow> </math></EquationSource> </InlineEquation> denotes the shares required for complete reconstruction. Comprehensive performance evaluation covering computational, storage, and communication costs demonstrates that the proposed approach effectively minimizes overhead while maintaining lossless reconstruction quality. Furthermore, extensive security analysis confirms that the scheme ensures information-theoretic secrecy, high share entropy, and strong resilience against collusion, noise, and cropping attacks. Experimental results validate that VS-MSS is efficient, secure, and scalable, making it well-suited for real-time and cloud-based multimedia applications.</p>

VS-MSS: Visual saliency-based efficient and secure multi-secret sharing scheme over cloud storage

期刊论文
发表日期：

2025年12月1日

查看全文

Arjun Singh Rawat,
Maroti Deshmukh,
Maheep Singh,
Sandeep Chand Kumain,
Lalit Kumar Awasthi

摘要

Existing state-of-the-art secret sharing schemes primarily focus on generating shares for the entire image. However, there is limited research in the literature that specifically focuses on objects or attention-worthy objects (salient objects) present in an image. Performing share generation and reconstruction over the complete image often leads to increased computational, communication, and storage overhead due to redundant background information. To overcome these limitations, this work proposes a visual saliency-based multi-secret sharing (VS-MSS) scheme for secure and efficient cloud storage, focusing exclusively on salient objects as the actual secrets. The proposed framework integrates a tri-stage encoder--decoder network for salient object detection with a lightweight arithmetic modulo-based secret sharing mechanism for efficient share generation and reconstruction. The method utilizes operations such as random permutation, modular arithmetic, and modular multiplicative inverse, achieving an optimal balance between data security and computational efficiency. The scheme follows a perfect $(n + 1,n + 1)$ $(n + 1, n + 1)$ multi-secret sharing configuration, where the first $n+1$ $n + 1$ represents the number of generated shares distributed across independent non-colluding cloud nodes, and the second $n+1$ $n + 1$ denotes the shares required for complete reconstruction. Comprehensive performance evaluation covering computational, storage, and communication costs demonstrates that the proposed approach effectively minimizes overhead while maintaining lossless reconstruction quality. Furthermore, extensive security analysis confirms that the scheme ensures information-theoretic secrecy, high share entropy, and strong resilience against collusion, noise, and cropping attacks. Experimental results validate that VS-MSS is efficient, secure, and scalable, making it well-suited for real-time and cloud-based multimedia applications.