.png){kind=icon}
.png){kind=icon}
Return
An efficient integrity-based video steganography framework using deep learning network
.png){kind=icon}
PRE
AI
DOI：10.1007/s11042-026-21430-3.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Secure video steganography is difficult to achieve because of the conflicting among payload capacity, visual imperceptibility and robustness. Existing approaches either produce distortions at high embedding rates or have unpredictable performance over different video contents. A novel hybrid video steganography framework based on integrity-encoded and GAN model is proposed. Huffman compression eliminates redundancy at the sender side while GAN-based feature selection is beneficial for nature image embedding. At the receiver side, message extraction is checked by discriminator-based integrity verification, which also guarantees that the message is not exposed to any potential adversaries during extraction. The proposed scheme achieves the reduction of the computational overheads to 39% compared with that of the GAN+CP-ABE models and the encryption and decryption time are 8934 ms and 8045 ms. Better accuracy and visual quality in different sizes of frame demonstrate that the method is applicable to the secure video communication.
Keywords:
Video Steganography
Transform Domain Embedding
DWT-based Data Hiding
Secure Message Extraction
Payload Optimization
Coefficient Modulation
Journal
IF:
3
Papers:
1.9W
Citations:
3.2W