Text Hiding by Anime Character Generation
(Encryption Decryption Framework)
Overview:
The privacy of secret information has gotten a lot of attention in the last few decades due to the newly developed hacking techniques and malware attacks. The image steganography played an important role in securely transmitting the secret information to the recipient. Image steganography is a process of integrating secret information into the carrier image by hiding the existence of secret information. Image steganography showed significant advancements in recent years after the rapid growth of Generative adversarial Networks (GAN). The majority of the work has been done by using the natural image as carrier images to hide secret information. The aim of the proposed work is to find the solution of text hiding, where the more secure and robust framework will generate the Stego image with increased payload capacity. The proposed framework will hide the text data by directly generating the anime characters without using a natural image (cover image). The anime characters dataset and news dataset will be used for the generation of Anime. Transfer learning of pre-trained GAN and CNN models will also be applied for the encryption and decryption of text data into anime characters.
Problem Statement
An encryption decryption-based text hiding framework to encrypt and decrypt text information with increased security and payload capacity. The framework will encrypt text by generating anime faces using GAN and decrypt anime faces by illustration module to extract text information.
Research Challenges/Gaps
Research Gap
Increase the payload Capacity of the stego image.
Maintain the security of the secret information with an increased payload capacity.
Development of a Robust Anime Generation model for the transmission of Secret Information in the Presence of different attacks with high accuracy.
Beneficiaries of the Project
Motivation
It will help businesses, security agencies, and large-scale applications to transfer their secret information without any unauthorized access.
Reduce the hackers attack by hiding the existence of secret information
It allows for copyright protection on digital files.
Objectives
Secure the communication of sensitive information between the two parties.
Bring comfort in sharing sensitive and secret information by generating anime characters without cover images.
Related Datasets
Danbooru 2019 portrait Faces
Proposed Architecture
Architecture of Proposed Framework
Stego Image Indexing
Indexing Algorithm
▸Generate a Grid of 8*8 using Secret Key.
▸Place Generated Stego image in Grid according to the position.
▸Place Random Images at 0 Indexes.
Results
Comparison of Results
SRGAN improve the results of approximately all face attributes.
Play a significant role in the robustness and security of the framework.
Hiding Capacity Evaluation
Encrypted bits per Anime Character = 17
Total Number of Anime Characters in Grid = 64
Total Encrypted bits = 64 * 17 = 1088
Robustness Evaluation
Clarity Evaluation
Security Evaluation
Numerous tools developed for the detection of stego images.
Evaluate the security of the proposed framework by Steg Analyzer.
Generated Images successfully deceive the Steg Analyzer [19] against detection.
Team
Supervisor
MS Scholar
Dr. Usama Ijaz Bajwa
Hafiz Abdul Rehman
usamabajwa@cuilahore.edu.pk
Co-PI, Video Analytics lab, National Centre in Big Data and Cloud Computing,
Program Chair (FIT 2019),
HEC Approved PhD Supervisor,
Associate Professor & Associate Head of Department
Department of Computer Science,
COMSATS University Islamabad, Lahore Campus, Pakistan
abdulrehman19427@gmail.com
Research Scholar (RCS),
Department of Computer Science,
COMSATS University Islamabad, Lahore Campus, Pakistan