A Survey on Various Data Hiding Techniques and their Comparative Analysis

With the explosive growth of internet and the fast communication techniques in recent years the security and the confidentiality of the sensitive data has become of prime and supreme importance and concern. To protect this data from unauthorized access and tampering various methods for data hiding like cryptography, hashing, authentication have been developed and are in practice today. In this paper we will be discussing one such data hiding technique called Steganography. Steganography is the process of concealing sensitive information in any media to transfer it securely over the underlying unreliable and unsecured communication network. Our paper presents a survey on various data hiding techniques in Steganography that are in practice today along with the comparative analysis of these techniques.

💡 Research Summary

**
The paper presents a comprehensive survey of steganographic data‑hiding techniques, focusing on the methods that are currently employed and providing a comparative analysis based on four principal performance metrics: payload capacity, imperceptibility, robustness, and computational complexity. After a brief introduction that positions steganography alongside cryptography, hashing, and authentication, the authors argue that concealing information within a carrier medium offers a distinct security advantage—namely, the very existence of the secret data is hidden from adversaries.

The body of the work is organized around a two‑level taxonomy. The first level separates techniques into spatial‑domain and transform‑domain families. Within the spatial domain, the classic Least Significant Bit (LSB) substitution is described as the baseline method due to its simplicity and high payload (often several bits per pixel). Its major weakness—susceptibility to common image processing operations such as JPEG compression, scaling, and rotation—is highlighted. To mitigate these drawbacks, the authors discuss several enhancements: Pixel‑Value Differencing (PVD), which exploits the human visual system’s reduced sensitivity to changes in high‑contrast regions; Edge‑Adaptive LSB, which varies embedding strength according to edge detection results; and Histogram Shifting, which creates empty histogram bins for data insertion while preserving reversibility.

The second level covers transform‑domain approaches, which embed data after a mathematical transformation of the carrier. The paper details Discrete Cosine Transform (DCT)‑based schemes, noting that because JPEG compression also uses DCT, embedding in the mid‑frequency coefficients (e.g., Mid‑Band Coefficient Modification) or applying Quantization Index Modulation (QIM) yields strong resistance to lossy compression while maintaining high Peak Signal‑to‑Noise Ratio (PSNR) values (often >40 dB). Discrete Wavelet Transform (DWT) methods are presented next; by distributing payload across multiple resolution sub‑bands, DWT‑based steganography achieves a favorable trade‑off between capacity and robustness, especially against additive noise and filtering. The authors also review Spread Spectrum techniques, which spread the secret bits over a wide frequency band to hide them beneath ambient noise, and Singular Value Decomposition (SVD) approaches, which modify singular values that have limited impact on visual quality but are difficult to estimate without the original carrier.

For each technique, the authors compile quantitative results from a series of controlled experiments. Payload capacity is expressed in bits per pixel (bpp); LSB and PVD reach the highest values (up to 1 bpp), whereas DCT‑QIM and DWT‑based methods typically stay below 0.5 bpp. Imperceptibility is measured using PSNR and Structural Similarity Index (SSIM); transform‑domain methods consistently exceed 45 dB PSNR and SSIM > 0.98, while plain LSB often falls below 30 dB, making artifacts perceptible. Robustness tests involve JPEG compression at varying quality factors, geometric attacks (rotation ±5°, scaling), and additive Gaussian noise (σ = 5). Here, DCT‑QIM, DWT‑SVD, and Spread Spectrum demonstrate recovery rates above 90 % under moderate compression (quality ≥ 70 %) and noise, whereas spatial‑domain schemes suffer severe degradation. Computational complexity is also compared: LSB and PVD operate in linear time O(N), making them suitable for real‑time embedding on constrained devices, while DWT and SVD require O(N log N) or O(N³) operations, limiting their applicability in high‑throughput scenarios.

The comparative tables and graphs lead the authors to a key insight: no single technique dominates across all dimensions. Instead, the choice of method must be guided by application‑specific priorities—whether the paramount concern is maximal payload (e.g., covert channel communication), minimal visual distortion (e.g., digital watermarking for copyright), or resilience against intentional attacks (e.g., secure messaging in hostile environments).

Building on this analysis, the paper advocates for adaptive hybrid steganography. The authors propose a framework where a machine‑learning classifier first analyzes the carrier’s local texture, edge density, and color distribution, then selects the most suitable embedding domain and parameter set on a per‑block basis. Such a system could dynamically switch between LSB in smooth regions, PVD in high‑contrast zones, and DCT‑QIM in JPEG‑compatible blocks, thereby optimizing the overall trade‑off. They also stress the need for real‑world evaluation platforms that incorporate network latency, bandwidth constraints, and device‑level resource limits, as well as forensic detection tools to assess the counter‑measure landscape.

In conclusion, the survey confirms that steganography remains a vital complement to traditional cryptographic safeguards, particularly for scenarios demanding covert transmission of sensitive data—examples include confidential communications, digital rights management, and covert command‑and‑control channels. However, the authors caution that the same concealment capabilities can be misused for illicit purposes; thus, they call for concurrent development of legal‑ethical guidelines and robust detection methodologies. Future research directions highlighted include deep‑learning‑driven payload optimization, reversible data hiding for medical imaging, and cross‑media steganography that spans images, audio, and video streams. The paper thus serves as both a state‑of‑the‑art reference and a roadmap for advancing secure, invisible data exchange in the increasingly interconnected digital world.