Deep Learning Based Coverless Video Steganography for Secret Audio
Hla Naing Win, Myo Khaing · 2025
Coverless video steganography has emerged as a promising technique for covert communication, offering a high level of security by avoiding direct modification of media content. In this research, a proposed system is implemented for Coverless Video Steganography that leverages Mel-Frequency Cepstral Coefficients (MFCC) and Convolutional Neural Network (CNN) features to securely embed secret audio information within video content without altering the cover media. This proposed method first extracts robust MFCC features from the secret audio, capturing perceptually relevant speech characteristics. These features are then mapped to semantically similar visual features within selected frames of a cover video through a CNN-based feature matching process, ensuring that the embedded information is not directly detectable. For enhanced security, the audio secret information is encrypted using RSA encryption before embedding, providing an additional layer of protection against unauthorized access. By integrating MFCC and CNN features with RSA encryption, this method effectively ensures the security and imperceptibility of secret data within the video. Experimental results demonstrate that the proposed method achieves high security, strong resilience against steganalysis attacks, and effective data retrieval accuracy, marking it as a viable solution for secure, undetectable audio communication. This proposed system applied for Secure Medical Record. In this system, embeds sensitive audio medical data, such as diagnostic results or doctor’s notes, into patient video files for secure and private storage or sharing with the better security performance.