Security Data Auditing based on Multifunction Digital Watermark for Multimedia File in Cloud Storage

Yongjun Ren, Jian Shen, Jin Wang, Jiang Xu, Liming FangInternational Journal of Multimedia and Ubiquitous Engineering 2014 · DOI

Abstract

Cloud computing is a promising computing model that enables convenient and on-demand network access to a shared pool of configurable computing resources. In cloud computing, data owners host their data on cloud servers and data consumers can access the data from cloud servers. This new paradigm of data storage service also introduces new security challenges, because data owners and data servers have different identities and different business interests. Therefore, an independent auditing service is required to make sure that the data is correctly hosted in the Cloud. However, the existing solutions are not specific to the multimedia data. Moreover, copyright protection is not provided.

Introduction

As data storage migrates to the cloud, a "trust but verify" dilemma emerges. Cloud Service Providers (CSPs) may be incentivized to delete rarely accessed data to optimize storage or conceal data loss due to infrastructure failures. To address this, researchers have developed Data Storage Auditing (DSA) and Provable Data Possession (PDP) schemes.

However, traditional PDP models often suffer from three major limitations when applied to multimedia files:

  1. Content Blindness: Most schemes treat data as a generic binary bitstream, ignoring the semantic meaning and specific characteristics of multimedia data.
  2. Communication Overhead: Verification often requires transferring significant amounts of metadata or "tags," increasing network strain.
  3. Lack of Copyright Protection: While existing tools can verify a file's presence, they do not provide mechanisms to prove ownership or prevent illegal propagation.

This paper proposes a specialized PDP model for multimedia files that utilizes a double-function, self-embedded digital watermark based on Singular Value Decomposition (SVD) to simultaneously audit integrity and ensure copyright protection.

Methodology

The proposed framework replaces traditional cryptographic tags with a digital watermarking process. This "self-embedding" approach involves several key steps:

  1. Block Division: The original image is partitioned into $n \times n$ non-overlapping sub-blocks (e.g., $8 \times 8$).
  2. SVD Transformation: Each block undergoes Singular Value Decomposition (SVD), a mathematical technique that extracts the most significant features of the pixel matrix.
  3. Watermark Generation: The system calculates the "norm" of the singular values. The parity of the highest-order bit of this norm generates a robust watermark bit ($W$).
  4. Self-Embedding: This watermark bit is embedded into the Least Significant Bits (LSB) of the image pixels.
  5. Verification: A Third-Party Auditor (TPA) can extract the watermark without the original image. For copyright verification, the extracted watermark is compared against the original using Normalized Correlation (NC). For integrity auditing, the system checks if the embedded bits still match the current singular value relationships of the image blocks.

Datasets

The researchers evaluated the method using a dataset of 50 standard test images ($512 \times 512$ pixels), including well-known benchmarks such as Lena, Baboon, Plane, and Boats.

Key Results and Discussion

The experimental results demonstrate that the proposed scheme achieves high invisibility and robustness. The Peak Signal-to-Noise Ratio (PSNR) for watermarked images was consistently greater than 35dB, the standard threshold for human invisibility.

Furthermore, the system showed significant resilience against common image processing attacks. As shown in the table below, the Normalized Correlation (NC) values remained high (typically between 0.96 and 0.99) even under Gaussian noise, salt-and-pepper noise, and cropping attacks. This high NC indicates that the watermark remains identifiable for copyright and auditing purposes even if the file is slightly degraded or partially modified.

Table 1: Performance metrics (NC and PSNR) under noise and cropping attacks for various test images.

Key Insights

  • The "Double Function" Concept: A single mechanism serves two goals—acting as both a tamper-evident seal for integrity and an owner's signature for copyright.
  • Blind Extraction: The auditing process does not require the original, unwatermarked image, which is essential for efficient cloud operations.
  • SVD Robustness: Basing the watermark on SVD norms makes the system resilient to minor transmission noise while remaining sensitive to actual data corruption.

Conclusion

While the current algorithm is optimized for images, it provides a robust foundation for multimedia cloud auditing. Future work could extend this approach to video or audio using different transformation techniques like Discrete Cosine Transform (DCT) or Discrete Wavelet Transform (DWT). Additionally, further research is needed to evaluate the system's performance against advanced AI-based tampering and sophisticated cryptographic attacks on LSB embedding.

Summary updated

Abstract

Cloud computing is a promising computing model that enables convenient and on-demand network access to a shared pool of configurable computing resources. In cloud computing, data owners host their data on cloud servers and data consumers can access the data from cloud servers. This new paradigm of data storage service also introduces new security challenges, because data owners and data servers have different identities and different business interests. Therefore, an independent auditing service is required to make sure that the data is correctly hosted in the Cloud. However, the existing solutions are not specific to the multimedia data. Moreover copyright protection is not provided.
BibTeX
@article{ren2014security,
  author = {Yongjun Ren and Jian Shen and Jin Wang and Jiang Xu and Liming Fang},
  title = {{Security Data Auditing based on Multifunction Digital Watermark for Multimedia File in Cloud Storage}},
  journal = {International Journal of Multimedia and Ubiquitous Engineering},
  year = {2014},
  doi = {10.14257/ijmue.2014.9.9.24},
}
Figure