Data Computing in Covert Domain

Zhenxing Qian, Zichi Wang, Xinpeng Zhang

This paper, "Computing in Covert Domain Using Data Hiding," introduces a fresh perspective on secure cloud computing. Rather than relying on the heavy mathematical machinery of encryption, the authors propose using steganography—the art of hiding information in plain sight—to perform private data processing.

TL;DR

The authors propose Data Computing in the Covert Domain (DCCD), a framework where sensitive data is hidden inside ordinary "cover" files (like images) before being sent to the cloud. This allows servers to perform basic arithmetic on the files without ever "seeing" the secret data, offering a significantly faster and more efficient alternative to traditional Homomorphic Encryption.

What Problem Does This Paper Address?

As we move more data to the cloud, privacy becomes a major bottleneck. We want the cloud to process our data (e.g., searching a database or calculating statistics) without the cloud provider actually seeing the content of that data.

The standard solution is Homomorphic Encryption (HE), which allows operations on ciphertexts that translate to operations on the underlying plaintexts. However, HE is notoriously slow, computationally expensive, and results in "ciphertext expansion," where the encrypted data is many times larger than the original. The authors identify a gap: we need a way to achieve secure outsourcing that is lightweight enough for mobile devices and common users.

Methods

The researchers introduce the DCCD framework, which shifts the computation from the "encrypted domain" to the "covert domain."

  1. Data Embedding: A sender takes two pieces of source data ($m_1, m_2$) and hides them inside two cover images ($X_1, X_2$) using a secret key $K$. They use a technique called matrix embedding in the Least Significant Bitplane (LSB). This essentially tweaks the low-level pixel data in a way that is invisible to the human eye.
  2. Cloud Computation: The resulting "stego" images ($Y_1, Y_2$) are uploaded to the server. The server performs a computation (like adding the images together) to produce a new result image ($Y'$). Crucially, the server has no idea it is processing secret data; it thinks it is just manipulating images.
  3. Result Extraction: The receiver gets $Y'$ and uses the secret key $K$ to extract the final result.

The paper demonstrates that because the embedding process is mathematically structured, operations performed on the stego images (addition, inner products, and outer products) map directly to the desired operations on the hidden source data.

Datasets

The authors evaluated their method using:

  • UCID (Uncompressed Colour Image Database): A standard benchmark containing 1,338 images used to test how well the data stays hidden.
  • Standard Test Images: Classic images like "Lena" and "Baboon" were used for specific feasibility demonstrations.

Key Insights and Ideas

  • Steganography as a Utility, Not Just Secrecy: Traditionally, steganography is used to send secret messages. This paper reimagines it as a tool for privacy-preserving outsourcing, where the "cover" acts as a natural, low-overhead container for computation.
  • Linearity is Key: The "aha moment" is realizing that LSB-based embedding can be treated as a linear operation. This allows the server to perform additive and multiplicative operations on the covers that "carry through" to the hidden payload.
  • Security through Obscurity + Cryptography: DCCD provides two layers of protection. First, an adversary might not even realize there is sensitive data to steal (imperceptibility). Second, even if they suspect data is there, they cannot extract the meaningful result without the secret key.

Key Results

  • Perfect Accuracy: The difference ratio between the calculated result and the extracted result was zero, meaning the covert computation is just as accurate as plaintext math.
  • High Imperceptibility: Using modern steganalysis tools (like SPAM and SRM), the authors showed that when hiding small amounts of data (under 1,000 bits), the existence of the secret data is virtually undetectable ($P_E$ values near 0.5).
  • Massive Speedup: The computational complexity of DCCD is significantly lower than RSA-based encryption. As the data size increases, the time required for DCCD stays relatively flat, whereas traditional encryption time climbs steeply.

Limitations and Future Work

While promising, the paper notes some clear boundaries:

  • Capacity Constraints: Unlike encryption, which can handle any amount of data, steganography is limited by the "size" of the cover. You can only hide so many bits in an image before it becomes visually distorted or detectable by algorithms.
  • Limited Operation Set: The paper focuses on addition and products. More complex non-linear functions (like those used in deep learning) would likely require more sophisticated embedding techniques.
  • Robustness: The current model assumes the server returns the image exactly as processed. If the server applies lossy compression (like JPEG) to the result, the hidden data might be corrupted. The authors suggest that future work could explore "robust" data hiding to solve this.

Summary updated

Abstract

This paper proposes an idea of data computing in the covert domain (DCCD). We show that with information hiding some data computing tasks can be executed beneath the covers like images, audios, random data, etc. In the proposed framework, a sender hides his source data into two covers and uploads them onto a server. The server executes computation within the stego and returns the covert computing result to a receiver. With the covert result, the receiver can extract the computing result of the source data. During the process, it is imperceptible for the server and the adversaries to obtain the source data as they are hidden in the cover. The transmission can be done over public channels. Meanwhile, since the computation is realized in the covert domain, the cloud cannot obtain the knowledge of the computing result. Therefore, the proposed idea is useful for cloud computing.
BibTeX
@inproceedings{qian2020data,
  author = {Zhenxing Qian and Zichi Wang and Xinpeng Zhang},
  title = {{Data Computing in Covert Domain}},
  booktitle = {SPDE},
  year = {2020},
  doi = {10.1007/978-981-15-9129-7\_37},
}

Loading related content...

Figure