[Paper Review] A Robust Data Hiding Process Contributing to the Development of a Semantic Web

Jacques M. Bahi, Jean-François Couchot+2 more|arXiv (Cornell University)|Jun 27, 2017

Advanced Steganography and Watermarking Techniques6 references3 citations

TL;DR

This paper proposes a robust steganographic scheme, $χDI_3$, based on chaotic iterations for embedding metadata into digital media to support the development of a Semantic Web. The method embeds tags in least significant coefficients using a significance function, achieving strong resistance to common attacks like cropping, JPEG/JPEG 2000 compression, and rotation, with error rates below 50% across all tested distortions.

ABSTRACT

In this paper, a novel steganographic scheme based on chaotic iterations is proposed. This research work takes place into the information hiding framework, and focus more specifically on robust steganography. Steganographic algorithms can participate in the development of a semantic web: medias being on the Internet can be enriched by information related to their contents, authors, etc., leading to better results for the search engines that can deal with such tags. As media can be modified by users for various reasons, it is preferable that these embedding tags can resist to changes resulting from some classical transformations as for example cropping, rotation, image conversion, and so on. This is why a new robust watermarking scheme for semantic search engines is proposed in this document. For the sake of completeness, the robustness of this scheme is finally compared to existing established algorithms.

Motivation & Objective

• To develop a steganographic scheme that enables persistent metadata embedding in digital media, supporting the vision of a Semantic Web where content is enriched with machine-readable descriptions.
• To ensure robustness of embedded tags against common user-induced transformations such as cropping, compression (JPEG, JPEG 2000), and rotation.
• To evaluate the performance of the proposed $χDI_3$ scheme in comparison with established steganographic algorithms, focusing on robustness and efficiency.
• To demonstrate the feasibility of using steganography as a foundation for decentralized, web-scale semantic tagging without relying on centralized databases or websites.

Proposed method

• The scheme uses a significance function $u^k = 8 - (k \mod 8)$ to identify least significant coefficients (LSCs) in pixel data, prioritizing bits with minimal perceptual impact.
• Embedding is performed via chaotic iterations, a deterministic yet unpredictable process based on a chaotic map, to map secret data into LSCs while preserving robustness.
• The payload is set at 0.1 bits per pixel, with a 262,144-bit message embedded into each $512 \times 512$ grayscale image from the BOSS database.
• Robustness is evaluated using controlled attacks: cropping (1% to 81%), JPEG and JPEG 2000 compression, and rotation (2° to 20°), with bit error rate (BER) computed as the percentage of mismatched bits.
• The method leverages the inherent sensitivity of chaotic systems to initial conditions to enhance security and resistance to reversal attacks.
• Experiments are conducted on 50 images from the BOSS contest, with results analyzed across multiple attack types to assess resilience.

Experimental results

Research questions

• RQ1 How effectively can the $χDI_3$ scheme embed metadata into digital media while maintaining perceptual transparency and robustness?
• RQ2 To what extent does $χDI_3$ resist common image transformations such as cropping, compression (JPEG and JPEG 2000), and rotation?
• RQ3 How does the robustness of $χDI_3$ compare to established steganographic schemes like YASS, nsF5, MMx, and HUGO under identical attack conditions?
• RQ4 Can the scheme support persistent metadata embedding in real-world web environments where media are frequently modified?
• RQ5 What is the trade-off between security, speed, and robustness in $χDI_3$ compared to existing topological steganographic schemes like $χCIS_2$?

Key findings

• The $χDI_3$ scheme demonstrates robustness against all tested attacks, with bit error rates consistently below 50% across all attack types.
• For JPEG compression, the scheme maintains a low error rate, indicating strong resilience to lossy compression commonly used in web media.
• JPEG 2000 compression also results in low error rates, confirming the scheme’s effectiveness in modern, high-efficiency compression formats.
• Cropping attacks up to 81% of the image size result in error rates below 50%, showing that the embedded data remains recoverable even after significant content removal.
• Rotation attacks of up to 20 degrees yield error rates below 50%, indicating that the scheme preserves embedded data under geometric transformations.
• The scheme is faster than the topologically secure $χCIS_2$ scheme, making it more suitable for real-time web applications requiring low latency.