Image Steganography With CNNs
Image Steganography With CNNs
with CNNs
This presentation explores image steganography using
Convolutional Neural Networks (CNNs). We will examine how
CNNs enhance the process of hiding information within
images.
Pooling Layers
Reduce dimensionality.
Activation Functions
Introduce non-linearity.
Loss Functions
Optimize the network.
Encoder Network
1 Embedding Role
Embeds secret image.
2 Encoder Input
Cover and secret images.
3 Convolution
Extract features for merging.
4 Stego Output
Creates the stego image.
Decoder Network
Extraction Role
Extracts the secret image.
Decoder Input
Receives the stego image.
Deconvolution
Reconstructs the secret image.
Secret Output
Recovers the original image.
ImageNet Dataset
Size 1 Diversity
2
Preprocessing
4 Suitability
3
ImageNet contains 1.2 million images. It has 1000 classes of objects. This diversity improves model
generalization.
Experimental Results
and Analysis
Advantages
2 Encoder-decoder benefits.
Improvements
3 Address limitations.
Future directions include video steganography. Also, explore adversarial attacks. Real-time applications
hold promise.