Detection-Aware ROI-Based Video Steganography Using Dual-Network Training

Abstract

This project proposes a detection-aware video steganography framework that integrates region-of-interest (ROI)-guided embedding with dual-network training. Objective: Improve imperceptibility and resistance to detection simultaneously Approach: Incorporate a steganalysis network during training for adaptive embedding Goal: Minimize detectability while preserving payload integrity and visual quality Hypothesis: Detection-aware training can reduce steganographic detectability while preserving visual quality. Significance: This work addresses vulnerabilities of modern steganography against deep learning-based detection.