Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Deep residual steganalysis method based on heterogeneous kernel

A steganalysis and heterogeneous core technology, applied in the field of content security, can solve the problems of limited image size, long training time, poor versatility, etc., to reduce interference, ensure accuracy, and improve versatility.

Active Publication Date: 2019-12-03
GUIZHOU NORMAL UNIVERSITY
View PDF2 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0002] Currently, steganalysis of convolutional neural networks faces pro

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Deep residual steganalysis method based on heterogeneous kernel
  • Deep residual steganalysis method based on heterogeneous kernel
  • Deep residual steganalysis method based on heterogeneous kernel

Examples

Experimental program
Comparison scheme
Effect test

Example Embodiment

[0034] The present invention will be further described below through the drawings and embodiments.

[0035] A deep residual steganalysis method based on heterogeneous cores. The SRM model is used in the preprocessing stage of ResNets, and the HetConv heterogeneous core is used as a convolution core in the DRHNet network. It includes the following steps:

[0036] Step 1. Extract the feature vector of each SRM sub-model of the image;

[0037] Step 2. Combine and reorganize all the feature vectors of the image;

[0038] Step 3. Put the original image and the encrypted image into the DRHNet network model for training;

[0039] Step 4. Put the image to be detected into the trained DRHNet steganalysis network for detection, and the result is used as a reference for whether the image is embedded.

[0040] Specifically, step 1. The SRM model uses k high-pass filters to extract the residuals of the overlay image or the hidden image to form k sub-models; then each sub-model is quantized, rounded, ...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a deep residual steganalysis method based on a heterogeneous kernel, and the method comprises the steps: introducing an SRM model at a ResNets preprocessing stage, and enablinga HetConv heterogeneous kernel to serve as a convolution kernel to be applied to a DRHNet network; the method further comprises the following steps: step 1, extracting a feature vector of each SRM sub-model of an image; step 2, combining and recombining all the feature vectors of the image; step 3, putting the original image and the secret-loaded image into a DRHNet network model for training; and step 4, putting the to-be-detected image into the trained DRHNet steganalysis network for detection, wherein a result is used as a reference for whether the image is embedded with secret keys. According to the method, while the steganalysis accuracy is ensured, the limitation of the size of the detected steganalysis image is overcome, the universality of the steganalysis network is improved, andthe training time of the network in the training stage is shortened.

Description

technical field [0001] The invention relates to the field of content security in cyberspace security, in particular to a deep residual steganalysis method based on heterogeneous kernels. Background technique [0002] Currently, steganalysis with convolutional neural networks faces problems such as poor generalizability, long training time, and limited image size. To address these issues, we propose a residual learning framework with heterogeneous kernels to save time during the training phase and improve the generality of steganalysis networks. As the input of the network, the present invention extracts and merges the image into the feature matrix through all sub-models of the Spatial Rich Model (SRM), and uses the 187×187 feature matrix as the actual input of the network. The architecture proposed by the present invention has good generality and can reduce the number of calculations and parameters, while still achieving higher accuracy. [0003] Convolutional Neural Netwo...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
IPC IPC(8): G06T1/00G06N3/04G06N3/08
CPCG06T1/0021G06N3/08G06N3/045
Inventor 徐洋付子爔许丹丹
Owner GUIZHOU NORMAL UNIVERSITY
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com