سامانه همایش های دانشگاه آزاد اسلامی واحد نجف آباد

صفحه اصلی / سومین کنفرانس بین المللی مهندسی و علوم کامپیوتر و اولین کنفرانس ملی سامانه های هوشمند، کلان داده و رایانش پیشرفته

A Comprehensive Survey and Comparative Analysis of Adaptive Reversible Data Hiding (RDH) in Audio: Time-Domain Signal Processing Techniques and Deep Learning Architectures for High Imperceptibility

نویسندگان :

Amir Lotfinia¹ Farzane Kaviani Baghbaderani²

1- دانشگاه آزاد خوراسگان 2- دانشگاه آزاد خوراسگان

کلمات کلیدی :

Reversible Data Hiding (RDH)،Audio Steganography،Invertible Neural Networks (INN)،Interpolation Sampling،Code Division Multiplexing (CDM)

چکیده :

Abstract—This review investigates the intricate engineering required to conceal data within audio signals, focusing specifically on the delicate "tug-of-war" inherent in Reversible Data Hiding (RDH) technologies. The fundamental challenge in modern steganographic design is managing a constant tension between three critical pillars: maximizing the data payload, ensuring the audio remains perceptually unchanged to the listener, and maintaining robustness so the message survives signal processing. While recent innovations like Code Division Multiplexing (CDM) have successfully scaled capacities to impressive rates—often exceeding 12 bits per sample or 2002 bits per second—engineers face a strict "fidelity ceiling," requiring a Signal-to-Noise Ratio (SNR) above 33 dB to prevent audible artifacts. Beyond simple concealment, the non-negotiable priority for forensic and legal applications is reversibility: the ability to extract the hidden data and then restore the cover audio to its pristine, bit-for-bit original state. Although time-domain approaches allow for direct manipulation without the phase distortion that plagues frequency-domain methods, they frequently struggle to mask "spectral artifacts," leaving distinct high-frequency signatures that make them vulnerable to advanced digital scanning. Consequently, this paper evaluates how emerging architectures, including Invertible Neural Networks (INNs) and adaptive expansion strategies, are attempting to resolve these long-standing conflicts to create communication channels that are both high-capacity and perfectly reversible.