SEAP: squeeze-and-excitation attention guided pruning for lightweight steganalysis networks
Abstract In recent years, the increasing computational and storage demands of deep steganalysis models have drawn attention to lightweight architectures. While pruning algorithms for image steganalysis networks have been proposed, they often do not apply to networks equipped with mobile inverted bot...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
SpringerOpen
2025-08-01
|
| Series: | EURASIP Journal on Information Security |
| Subjects: | |
| Online Access: | https://doi.org/10.1186/s13635-025-00212-8 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849342553387696128 |
|---|---|
| author | Qiushi Li Shenghai Luo Shunquan Tan Zhenjun Li |
| author_facet | Qiushi Li Shenghai Luo Shunquan Tan Zhenjun Li |
| author_sort | Qiushi Li |
| collection | DOAJ |
| description | Abstract In recent years, the increasing computational and storage demands of deep steganalysis models have drawn attention to lightweight architectures. While pruning algorithms for image steganalysis networks have been proposed, they often do not apply to networks equipped with mobile inverted bottleneck (MBConv) structures, such as EfficientNet. In this paper, we propose a Squeeze-and-Excitation Attention-based Pruning framework for image steganalysis networks, named SEAP. The method adopts a block-wise structured pruning strategy guided by the SE channel attention mechanism, where unimportant channels within each MBConv block are identified based on SE attention values and soft masks. Since pruning is conducted independently within each MBConv block and the input/output dimensions of the block remain unchanged, potential pruning conflicts across blocks are effectively avoided. In addition, we propose a sparsity regularization mechanism that adaptively adjusts the regularization strength based on the network structure, helping to preserve detection performance. Extensive experimental results demonstrate that the pruned network retains only a small fraction of the original network’s parameters and computational costs while achieving performance comparable to the original unpruned networks. |
| format | Article |
| id | doaj-art-d16078da112944b5bcfb1762a64f64cd |
| institution | Kabale University |
| issn | 2510-523X |
| language | English |
| publishDate | 2025-08-01 |
| publisher | SpringerOpen |
| record_format | Article |
| series | EURASIP Journal on Information Security |
| spelling | doaj-art-d16078da112944b5bcfb1762a64f64cd2025-08-20T03:43:21ZengSpringerOpenEURASIP Journal on Information Security2510-523X2025-08-012025111510.1186/s13635-025-00212-8SEAP: squeeze-and-excitation attention guided pruning for lightweight steganalysis networksQiushi Li0Shenghai Luo1Shunquan Tan2Zhenjun Li3Guangdong Laboratory of Machine Perception and Intelligent Computing, Faculty of Engineering, Shenzhen MSU-BIT UniversityGuangdong Laboratory of Machine Perception and Intelligent Computing, Faculty of Engineering, Shenzhen MSU-BIT UniversitySchool of Information and Communications Technology, Shenzhen City PolytechnicSchool of Information and Communications Technology, Shenzhen City PolytechnicAbstract In recent years, the increasing computational and storage demands of deep steganalysis models have drawn attention to lightweight architectures. While pruning algorithms for image steganalysis networks have been proposed, they often do not apply to networks equipped with mobile inverted bottleneck (MBConv) structures, such as EfficientNet. In this paper, we propose a Squeeze-and-Excitation Attention-based Pruning framework for image steganalysis networks, named SEAP. The method adopts a block-wise structured pruning strategy guided by the SE channel attention mechanism, where unimportant channels within each MBConv block are identified based on SE attention values and soft masks. Since pruning is conducted independently within each MBConv block and the input/output dimensions of the block remain unchanged, potential pruning conflicts across blocks are effectively avoided. In addition, we propose a sparsity regularization mechanism that adaptively adjusts the regularization strength based on the network structure, helping to preserve detection performance. Extensive experimental results demonstrate that the pruned network retains only a small fraction of the original network’s parameters and computational costs while achieving performance comparable to the original unpruned networks.https://doi.org/10.1186/s13635-025-00212-8SteganalysisSteganographyDeep learningConvolutional neural networkNetwork pruning |
| spellingShingle | Qiushi Li Shenghai Luo Shunquan Tan Zhenjun Li SEAP: squeeze-and-excitation attention guided pruning for lightweight steganalysis networks EURASIP Journal on Information Security Steganalysis Steganography Deep learning Convolutional neural network Network pruning |
| title | SEAP: squeeze-and-excitation attention guided pruning for lightweight steganalysis networks |
| title_full | SEAP: squeeze-and-excitation attention guided pruning for lightweight steganalysis networks |
| title_fullStr | SEAP: squeeze-and-excitation attention guided pruning for lightweight steganalysis networks |
| title_full_unstemmed | SEAP: squeeze-and-excitation attention guided pruning for lightweight steganalysis networks |
| title_short | SEAP: squeeze-and-excitation attention guided pruning for lightweight steganalysis networks |
| title_sort | seap squeeze and excitation attention guided pruning for lightweight steganalysis networks |
| topic | Steganalysis Steganography Deep learning Convolutional neural network Network pruning |
| url | https://doi.org/10.1186/s13635-025-00212-8 |
| work_keys_str_mv | AT qiushili seapsqueezeandexcitationattentionguidedpruningforlightweightsteganalysisnetworks AT shenghailuo seapsqueezeandexcitationattentionguidedpruningforlightweightsteganalysisnetworks AT shunquantan seapsqueezeandexcitationattentionguidedpruningforlightweightsteganalysisnetworks AT zhenjunli seapsqueezeandexcitationattentionguidedpruningforlightweightsteganalysisnetworks |