STCL:Curriculum learning Strategies for deep learning image steganography models

TL;DR

STCL addresses the dual challenges of image quality and slow convergence in deep-learning image steganography by introducing a curriculum-learning framework guided by difficulty estimates from multiple teacher models and a knee-point based training schedule. It partitions training data into Easy, Medium, and Difficult subsets using per-sample quality consistency (measured by $S_{ij}$ and $P_{ij}$ across teacher levels) and trains in staged phases that stop at knee points to avoid overfitting and speed up convergence. Empirical results on ALASKA2, VOC2012, and ImageNet show consistent improvements in $PSNR$, $SSIM$, MSSSIM, and decoding accuracy, along with lower steganalysis scores, and demonstrate generalization to other architectures. The approach offers a practical route to higher-quality, more secure steganographic systems with improved training efficiency, while leaving room to extend to non-binary embedding in future work.

Abstract

Aiming at the problems of poor quality of steganographic images and slow network convergence of image steganography models based on deep learning, this paper proposes a Steganography Curriculum Learning training strategy (STCL) for deep learning image steganography models. So that only easy images are selected for training when the model has poor fitting ability at the initial stage, and gradually expand to more difficult images, the strategy includes a difficulty evaluation strategy based on the teacher model and an knee point-based training scheduling strategy. Firstly, multiple teacher models are trained, and the consistency of the quality of steganographic images under multiple teacher models is used as the difficulty score to construct the training subsets from easy to difficult. Secondly, a training control strategy based on knee points is proposed to reduce the possibility of overfitting on small training sets and accelerate the training process. Experimental results on three large public datasets, ALASKA2, VOC2012 and ImageNet, show that the proposed image steganography scheme is able to improve the model performance under multiple algorithmic frameworks, which not only has a high PSNR, SSIM score, and decoding accuracy, but also the steganographic images generated by the model under the training of the STCL strategy have a low steganography analysis scores. You can find our code at \href{https://github.com/chaos-boops/STCL}{https://github.com/chaos-boops/STCL}.

Figures (8)

• Figure 1: Difficulty Evaluation Strategies Based on Teacher Models.
• Figure 2: Training subsets with different difficulties obtained by the teacher model difficulty evaluation method.
• Figure 3: Multi-stage scheduling rules based on knee points.
• Figure 4: Comparison of cover and stego images under 1-3 bpp capacity steganography.
• Figure 5: Comparison of stego image generated by multi-stage model and cover image.