Strategy to improve the accuracy of convolutional neural network architectures applied to digital image steganalysis in the spatial domain
| Autor: | Gustavo Isaza, Simon Orozco-Arias, Reinel Tabares-Soto, Mario Alejandro Bravo-Ortiz, Raul Ramos Pollan, Daniel Arias-Garzón, Alejandro Burbano Jacome, Alejandro Mora-Rubio, Harold Brayan Arteaga-Arteaga, Jesus Alejandro Alzate Grisales |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
General Computer Science
Computer science Computer Vision Feature extraction Strategy 0211 other engineering and technologies Stability (learning theory) Normalization (image processing) Convolutional neural network 02 engineering and technology lcsh:QA75.5-76.95 Artificial Intelligence 0202 electrical engineering electronic engineering information engineering Dropout (neural networks) Steganalysis 021110 strategic defence & security studies Contextual image classification business.industry Deep learning Security and Privacy Pattern recognition Cryptography 020201 artificial intelligence & image processing lcsh:Electronic computers. Computer science Artificial intelligence business |
| Zdroj: | PeerJ Computer Science, Vol 7, p e451 (2021) PeerJ Computer Science |
| ISSN: | 2376-5992 |
| DOI: | 10.7717/peerj-cs.451 |
| Popis: | In recent years, Deep Learning techniques applied to steganalysis have surpassed the traditional two-stage approach by unifying feature extraction and classification in a single model, the Convolutional Neural Network (CNN). Several CNN architectures have been proposed to solve this task, improving steganographic images’ detection accuracy, but it is unclear which computational elements are relevant. Here we present a strategy to improve accuracy, convergence, and stability during training. The strategy involves a preprocessing stage with Spatial Rich Models filters, Spatial Dropout, Absolute Value layer, and Batch Normalization. Using the strategy improves the performance of three steganalysis CNNs and two image classification CNNs by enhancing the accuracy from 2% up to 10% while reducing the training time to less than 6 h and improving the networks’ stability. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|