# CALPA-NET: Channel-pruning-assisted Deep Residual Network for Steganalysis of Digital Images

arXiv: https://arxiv.org/abs/1911.04657 · HTML: https://mlsanity.com/paper/1911.04657/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 12 Nov 2019.

## Built on by (most cited first, 6 of 6)

- [Deep Learning for Steganalysis of Diverse Data Types: A review of methods, taxonomy, challenges and future directions](https://mlsanity.com/paper/2308.04522/) (2023)
- [MCTSteg: A Monte Carlo Tree Search-based Reinforcement Learning Framework for Universal Non-additive Steganography](https://mlsanity.com/paper/2103.13689/) (2021)
- [Unified High-binding Watermark for Unconditional Image Generation Models](https://mlsanity.com/paper/2310.09479/) (2023)
- [Stego Battlefield: Evaluating Image Steganography Attacks and Steganalysis Defenses](https://mlsanity.com/paper/2605.05789/) (2026)
- [STD-NET: Search of Image Steganalytic Deep-learning Architecture via Hierarchical Tensor Decomposition](https://mlsanity.com/paper/2206.05651/) (2022)
- [Natias: Neuron Attribution based Transferable Image Adversarial Steganography](https://mlsanity.com/paper/2409.04968/) (2024)

## Abstract (arXiv)

Over the past few years, detection performance improvements of deep-learning based steganalyzers have been usually achieved through structure expansion. However, excessive expanded structure results in huge computational cost, storage overheads, and consequently difficulty in training and deployment. In this paper we propose CALPA-NET, a ChAnneL-Pruning-Assisted deep residual network architecture search approach to shrink the network structure of existing vast, over-parameterized deep-learning based steganalyzers. We observe that the broad inverted-pyramid structure of existing deep-learning based steganalyzers might contradict the well-established model diversity oriented philosophy, and therefore is not suitable for steganalysis. Then a hybrid criterion combined with two network pruning schemes is introduced to adaptively shrink every involved convolutional layer in a data-driven manner. The resulting network architecture presents a slender bottleneck-like structure. We have conducted extensive experiments on BOSSBase+BOWS2 dataset, more diverse ALASKA dataset and even a large-scale subset extracted from ImageNet CLS-LOC dataset. The experimental results show that the model structure generated by our proposed CALPA-NET can achieve comparative performance with less than two percent of parameters and about one third FLOPs compared to the original steganalytic model. The new model possesses even better adaptivity, transferability, and scalability.