Efficientnets Aplicadas à Esteganálise Em Imagens Digitais
Autor(a) principal: | |
---|---|
Data de Publicação: | 2022 |
Outros Autores: | , , , , |
Tipo de documento: | Artigo |
Idioma: | por |
Título da fonte: | Revista de Engenharia e Pesquisa Aplicada |
Texto Completo: | http://revistas.poli.br/index.php/repa/article/view/2215 |
Resumo: | Several CNN architectures with a specific purpose for steganalysis were developed and reached the state-of-the-art, surpassing the previous models that were based on the feature extraction and classification steps. New image datasets were proposed, differing from the previous ones by the number of instances and the variation of important characteristics such as the quality factor and the payload of hidden messages between images. In addition, new general-purpose architectures are applicable in the scope of steganalysis and benefit from transfer learning to accelerate training. This work presents the training of the Seteganalys Residual Network (SRNET) with random initialization of weights and performs the performance comparison between the CNN architectures Efficientnet and Efficientnetv2, with the latter perfoming 32% faster than EfficientnetB4, for each training epoch. Finally, an experiment involving successive training within the cover image and its respective stego-images. |
id |
UFPE-2_09bf4d7a68a163458104fd3d977df94c |
---|---|
oai_identifier_str |
oai:ojs.poli.br:article/2215 |
network_acronym_str |
UFPE-2 |
network_name_str |
Revista de Engenharia e Pesquisa Aplicada |
repository_id_str |
|
spelling |
Efficientnets Aplicadas à Esteganálise Em Imagens DigitaisSeveral CNN architectures with a specific purpose for steganalysis were developed and reached the state-of-the-art, surpassing the previous models that were based on the feature extraction and classification steps. New image datasets were proposed, differing from the previous ones by the number of instances and the variation of important characteristics such as the quality factor and the payload of hidden messages between images. In addition, new general-purpose architectures are applicable in the scope of steganalysis and benefit from transfer learning to accelerate training. This work presents the training of the Seteganalys Residual Network (SRNET) with random initialization of weights and performs the performance comparison between the CNN architectures Efficientnet and Efficientnetv2, with the latter perfoming 32% faster than EfficientnetB4, for each training epoch. Finally, an experiment involving successive training within the cover image and its respective stego-images.Diversas arquiteturas CNN com propósito específico para esteganálise foram desenvolvidas e atingiram o estado-da-arte superando os modelos anteriores que se baseavam nas etapas de extração de características e classificação. Novos conjuntos de dados de imagens foram propostos diferenciando-se dos anteriores pela quantidade de instâncias e a variação de características importantes como fator de qualidade e a carga útil (payload) de mensagem escondida em imagens. Além disso, novas arquiteturas de propósito geral têm se mostrado aplicáveis no âmbito da esteganálise e se beneficiam de transfer learning para acelerar o treinamento. Este trabalho aborda o treinamento da Steganalysis Residual Network (SRNET) com inicialização aleatória dos pesos e realiza a comparação de desempenho entre as arquiteturas de CNN Efficientnet e Efficientnetv2, com este último sendo 32% mais rápido que a EfficientnetB4, para cada época de treinamento. Por fim, também é apresentado um experimento envolvendo treinamentos sucessivos entre a imagem de cobertura e suas respectivas estego-imagens. Escola Politécnica de Pernambuco2022-07-15info:eu-repo/semantics/articleinfo:eu-repo/semantics/publishedVersionAvaliado pelos paresapplication/pdftext/htmlhttp://revistas.poli.br/index.php/repa/article/view/221510.25286/repa.v7i2.2215Journal of Engineering and Applied Research; Vol 7 No 2 (2022): Edição Especial em Inteligência Artificial; 32-41Revista de Engenharia e Pesquisa Aplicada; v. 7 n. 2 (2022): Edição Especial em Inteligência Artificial; 32-412525-425110.25286/repa.v7i2reponame:Revista de Engenharia e Pesquisa Aplicadainstname:Universidade Federal de Pernambuco (UFPE)instacron:UFPEporhttp://revistas.poli.br/index.php/repa/article/view/2215/817http://revistas.poli.br/index.php/repa/article/view/2215/818-Copyright (c) 2022 Thaise dos Santos Tenório, Rafael Albuquerque, Arlington Rodrigues, Gildo Ferrucio, Julia Aguiar, José Amarildo Filho, Francisco Madeirohttp://creativecommons.org/licenses/by-nc/4.0info:eu-repo/semantics/openAccessAlbuquerque, RafaelRodrigues, ArlingtonFerrucio, GildoAguiar, JuliaFilho, José AmarildoMadeiro, Francisco2022-07-17T20:06:52Zoai:ojs.poli.br:article/2215Revistahttp://revistas.poli.br/index.php/repaONGhttp://revistas.poli.br/index.php/repa/oai||repa@poli.br2525-42512525-4251opendoar:2022-07-17T20:06:52Revista de Engenharia e Pesquisa Aplicada - Universidade Federal de Pernambuco (UFPE)false |
dc.title.none.fl_str_mv |
Efficientnets Aplicadas à Esteganálise Em Imagens Digitais |
title |
Efficientnets Aplicadas à Esteganálise Em Imagens Digitais |
spellingShingle |
Efficientnets Aplicadas à Esteganálise Em Imagens Digitais Albuquerque, Rafael |
title_short |
Efficientnets Aplicadas à Esteganálise Em Imagens Digitais |
title_full |
Efficientnets Aplicadas à Esteganálise Em Imagens Digitais |
title_fullStr |
Efficientnets Aplicadas à Esteganálise Em Imagens Digitais |
title_full_unstemmed |
Efficientnets Aplicadas à Esteganálise Em Imagens Digitais |
title_sort |
Efficientnets Aplicadas à Esteganálise Em Imagens Digitais |
author |
Albuquerque, Rafael |
author_facet |
Albuquerque, Rafael Rodrigues, Arlington Ferrucio, Gildo Aguiar, Julia Filho, José Amarildo Madeiro, Francisco |
author_role |
author |
author2 |
Rodrigues, Arlington Ferrucio, Gildo Aguiar, Julia Filho, José Amarildo Madeiro, Francisco |
author2_role |
author author author author author |
dc.contributor.author.fl_str_mv |
Albuquerque, Rafael Rodrigues, Arlington Ferrucio, Gildo Aguiar, Julia Filho, José Amarildo Madeiro, Francisco |
description |
Several CNN architectures with a specific purpose for steganalysis were developed and reached the state-of-the-art, surpassing the previous models that were based on the feature extraction and classification steps. New image datasets were proposed, differing from the previous ones by the number of instances and the variation of important characteristics such as the quality factor and the payload of hidden messages between images. In addition, new general-purpose architectures are applicable in the scope of steganalysis and benefit from transfer learning to accelerate training. This work presents the training of the Seteganalys Residual Network (SRNET) with random initialization of weights and performs the performance comparison between the CNN architectures Efficientnet and Efficientnetv2, with the latter perfoming 32% faster than EfficientnetB4, for each training epoch. Finally, an experiment involving successive training within the cover image and its respective stego-images. |
publishDate |
2022 |
dc.date.none.fl_str_mv |
2022-07-15 |
dc.type.driver.fl_str_mv |
info:eu-repo/semantics/article info:eu-repo/semantics/publishedVersion Avaliado pelos pares |
format |
article |
status_str |
publishedVersion |
dc.identifier.uri.fl_str_mv |
http://revistas.poli.br/index.php/repa/article/view/2215 10.25286/repa.v7i2.2215 |
url |
http://revistas.poli.br/index.php/repa/article/view/2215 |
identifier_str_mv |
10.25286/repa.v7i2.2215 |
dc.language.iso.fl_str_mv |
por |
language |
por |
dc.relation.none.fl_str_mv |
http://revistas.poli.br/index.php/repa/article/view/2215/817 http://revistas.poli.br/index.php/repa/article/view/2215/818 |
dc.rights.driver.fl_str_mv |
http://creativecommons.org/licenses/by-nc/4.0 info:eu-repo/semantics/openAccess |
rights_invalid_str_mv |
http://creativecommons.org/licenses/by-nc/4.0 |
eu_rights_str_mv |
openAccess |
dc.format.none.fl_str_mv |
application/pdf text/html |
dc.coverage.none.fl_str_mv |
- |
dc.publisher.none.fl_str_mv |
Escola Politécnica de Pernambuco |
publisher.none.fl_str_mv |
Escola Politécnica de Pernambuco |
dc.source.none.fl_str_mv |
Journal of Engineering and Applied Research; Vol 7 No 2 (2022): Edição Especial em Inteligência Artificial; 32-41 Revista de Engenharia e Pesquisa Aplicada; v. 7 n. 2 (2022): Edição Especial em Inteligência Artificial; 32-41 2525-4251 10.25286/repa.v7i2 reponame:Revista de Engenharia e Pesquisa Aplicada instname:Universidade Federal de Pernambuco (UFPE) instacron:UFPE |
instname_str |
Universidade Federal de Pernambuco (UFPE) |
instacron_str |
UFPE |
institution |
UFPE |
reponame_str |
Revista de Engenharia e Pesquisa Aplicada |
collection |
Revista de Engenharia e Pesquisa Aplicada |
repository.name.fl_str_mv |
Revista de Engenharia e Pesquisa Aplicada - Universidade Federal de Pernambuco (UFPE) |
repository.mail.fl_str_mv |
||repa@poli.br |
_version_ |
1798036000445825024 |