Please enable JavaScript.
Coggle requires JavaScript to display documents.
Coverless information hiding based on the generation of anime characters -…
Coverless information hiding based on the generation of anime characters
Steganalysis
Information hiding method to embed secret information into the carrier in an invisible way
However, it is possible for steganalysis algorithms to detect the existence of hidden behavior
coverless information hiding (CIH)
Directly generate
the stego-carriers
media
sound
video
images
text
Retrieve
image selection methods
to establish a large-scale natural image database
retrieve the natural images that can express secret information as stego-images
hiding capacity is very limited
rely on image retrieval
information hiding based on carrier selection
transmit secret information by image hash sequence
Hard to apply them in real-world scenarios
set of sub-image blocks at specific locations of natural images
gray gradient co-occurrence matrix to encode the image
dynamic content selection framework applied to CIH
semi-creative methods
texture synthesis
do not need to establish a natural image database in advance
can synthesize an arbitrary size stego-image according to agreed rules
it has a large hidden capacity
there are quilting between the samples
the possibility of detection
Generating the stego-carrier according to agreed rules without specifying the original carrier in advance
stego-carrier
belongs
texture images
used code: LBP
Local Binary Pattern
to establish the mapping relationship between
3 more items...
building a basic sample database in advance
selecting the appropriate samples according to secret information
synthesize a relatively natural large stego-image according to agreed rules
MNIST data set
Modified National Institute of Standards and Technology database
to represent secret information
can clearly see the category label of secret information encoding
has lower security
Solutions
image repair technology
Cardan Grille Mask (CGK)
Labels “1~9”
creative methods
generative adversarial networks (GANs)
stego-image
anime character images are more resistant to the computer and visual detection
maintaining the consistency of statistical characteristics with the training image set
Achieving the degree of visual effect that the fake can be confused with the real
Secret information
into a collection of attribute labels for anime characters
hairstyle
hair color
eye color
etc
Stego-image
N × N
anime characters are combined in
row
column
0
1 more item...
other numbers
1 more item...
1
1 more item...
Robustness
Rotational attack
with angles of
3 more items...
Scaling
with scaling ratios of
4 more items...
Gaussian noise
mean value is 0
variance is
3 more items...
salt and pepper noise
mean value of noise is 0
the variance is
1 more item...
Speckle noise
Median filtering
Template sizes
1 more item...
Mean filtering
gaussian filtering
Information hiding
The steganography process
Step 1
Convert the secret information into a binary string
1 more item...
Step 2
Convert secret information sequence into anime character attribute label sequence L
1 more item...
Step 3
generate the anime character expressing secret information at the index position of the stego-image
1 more item...
Step 4
Send the generated stego-image to the receiver
to hide
numbers
specific images
Summary
Using the attributes of anime characters to represent secret information
the GAN is adopted to generate stego-images directly according
to the attributes of anime characters.
There is no necessary connection between the secret information and the anime character generation network.
with the existing GAN-based methods of CIH, the proposed method can transmit arbitrary secret information
Proposed CIH method
following three modules
secret information and label set conversion module (LSTM)
Illustration2Vec
network used to predict the attribute labels of anime characters
For instance, for a given image, Illustration2Vec can predict the probability of 512 attributes
For this situation
the attributes with a high probability of occurrence
5 hairstyles
2 bit (for 1 hairstyle)
13 hair colors
3 bit (for 1 hair color)
10 eye colors
3 bit (for 1 eye color)
other attributes
“blush”
1 more item...
“smile”
1 more item...
“open mouth”
1 more item...
“ribbon”
1 more item...
long short-term memory network (LSTM) is used
to convert
secret information
labels set
can generate high-quality text content
can predict the probability of the next output content in the case of existing partial input.
is trained
using database of a lot of anime characters from Getchu
2 more items...
Process
Step1
Convert the secret information into a binary string
1 more item...
Step 2
Generate the first attribute mapping through a pseudo-random transformation function
1 more item...
Step 3
Take the first attribute label l1 as input and predict the next label through the LSTM network
1 more item...
Step 4
the probability of the predicted nth attribute can be expressed as p(ln| l1, l2, …, ln − 1)
1 more item...
Step 5
Repeat Step 4 until the combination of attributes representing an anime characters is completed
Comparison
has the same clarity as the image in the training se
can express secret information correctly
to carry out label extraction
location index module
secret image generation module
Process
sender first converts the secret information into the corresponding attribute labels
generates anime characters that can express secret information according to the attribute labels at a specific index position
receiver first extracts the attribute labels of the anime characters at the index position
converts it into secret information
Source
https://jivp-eurasipjournals.springeropen.com/articles/10.1186/s13640-020-00524-4
Definitions
Steganalysis, CIH (Complex Instruction Set Computer), GANs (Generative Adversarial Networks), stego-image, LSTM (Long Short-Term Memory), steganography, robustness, pseudo-random, co-occurrence, framework
The process of detecting and analyzing hidden messages, data, or information concealed within digital media using steganography techniques.
A reusable and customizable structure or set of tools, libraries, and conventions that provide a foundation for developing software applications.
Refers to a sequence of numbers or data that appears to be random but is generated by a deterministic process
Refers to a type of computer architecture where the instruction set includes numerous complex instructions that can perform multiple operations in a single instruction.
An image that has been altered or manipulated using steganography techniques to hide secret information.
The practice of concealing secret information within other non-secret data, such as images, audio files, or text, in order to hide the existence of the secret information.
Refers to the simultaneous or sequential appearance of two or more entities or events within a certain context or timeframe.
Refers to the ability of a system or algorithm to maintain its performance or functionality under varying conditions or in the presence of disturbances or uncertainties.
A type of recurrent neural network (RNN) architecture designed to overcome the vanishing gradient problem in traditional RNNs.
A class of artificial intelligence algorithms used in unsupervised machine learning.
