Abstract:Although significant progress has been achieved recently in automatic learning of steganographic cost, the existing methods designed for spatial images cannot be directly applied to JPEG images which are more common media in daily life. The difficulties of migration are mainly caused by the characteristics of the $8\times 8$ DCT mode structure. To address the issue, in this paper we extend an existing automatic cost learning scheme to JPEG, where the proposed scheme called JEC-RL (JPEG Embedding Cost with Reinforcement Learning) is explicitly designed to tailor the JPEG DCT structure. It works with the embedding action sampling mechanism under reinforcement learning, where a policy network learns the optimal embedding policies via maximizing the rewards provided by an environment network. Following a domain-transition design paradigm, the policy network is composed of three modules, i.e., pixel-level texture complexity evaluation module, DCT feature extraction module, and mode-wise rearrangement module. These modules operate in serial, gradually extracting useful features from a decompressed JPEG image and converting them into embedding policies for DCT elements, while considering JPEG characteristics including inter-block and intra-block correlations simultaneously. The environment network is designed in a gradient-oriented way to provide stable reward values by using a wide architecture equipped with a fixed preprocessing layer with $8\times 8$ DCT basis filters. Extensive experiments and ablation studies demonstrate that the proposed method can achieve good security performance for JPEG images against both advanced feature-based and modern CNN-based steganalyzers.

Non-Additive Cost Functions for JPEG Steganography Based on Block Boundary Maintenance

A New Steganographic Method Resisting Statistical Attack

Steganalytic Scheme for JPEG Images Based on Multivariate Regression

Jpeg Quantization-Distribution Steganalytic Method Attacking Jsteg

A Dct-Based Image Steganographic Method Resisting Statistical Attacks

BBC++: Enhanced Block Boundary Continuity on Defining Non-Additive Distortion for JPEG Steganography.

Blind Jpeg Steganalysis Using Features Derived from Multi-Domain

Defining Joint Distortion for JPEG Steganography.

Controversial 'pixel' Prior Rule for JPEG Adaptive Steganography

A co-occurrence matrix based approach to detect jpeg steganography

Asymmetric Distortion Function for JPEG Steganography Using Block Artifact Compensation.

Non-Additive Cost Functions for Color Image Steganography Based on Inter-Channel Correlations and Differences

A New Cost Function for Spatial Image Steganography Based on 2D-SSA and WMF

Improving Cost Learning for JPEG Steganography by Exploiting JPEG Domain Knowledge

Domain Transformation of Distortion Costs for Efficient JPEG Steganography with Symmetric Embedding

Based the Magnitude of DCT Coefficient Adaptive Steganographic Method for JPEG Images

Defining Cost Functions for Adaptive Steganography at the Microscale.

Distortion Function Based on Residual Blocks for JPEG Steganography

Robust Steganography with Boundary-Preserving Overflow Alleviation and Adaptive Error Correction

On Improving Distortion Functions for JPEG Steganography.

A New Rule for Cost Reassignment in Adaptive Steganography.