Loading...

Steganalysis of JPEG Images Using Convolutional Neural Network

Sargazi Moghadam, Mohammad Hadi | 2020

416 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 52727 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Ghaemmaghami, Shahrokh
  7. Abstract:
  8. Steganalysis is the process of discovering existence of hidden data in a file or image. Recent advances in steganographic methods result in their wide use in secure communications applications. Reports of using information hiding by criminals to send messages, and malwares to communicate to servers, emphasizes importance of steganalysis in real time applications. Due to wide use of JPEG images in social media and internet, steganalysis of JPEG images is a highly active research topic in information hiding. Number of steganographic tools available for steganography in JPEG domain, shows importance of steganalysis of this image format. Classic Steganalysis techniques combine feature extraction methods and a classifier to perform this task. These techniques usually suffer from low accuracy in low embedding rates, which is one of the main problems in steganalysis of images. Although modern steganalysis techniques based on deep neural networks, improve detection accuracy of classic feature based methods, their computational complexity is a severe problem in real time applications. In this thesis, a method based on convolutional neural network is presented to improve steganalysis of JPEG images. A preprocessing block splits input image into multiple blocks to feed a convolutional neural network. Structure of network layers is properly designed to distinguish between embedded and original blocks. Finally, classification of images is performed using block labels. This block based technique is the main contribution of this steganalysis method. High detection accuracy and low computation complexity are main results of this research. Simulations show that detection accuracy of this method for PQ embedding is 98.8% which outperforms results obtained in feature based steganalysis (84.45%) and neural network steganalysis (98.36%). Also a convolutional neural network is proposed as an improvement of recent methods for steganalysis of HUGO embedding that results in 98.45% accuracy which outperforms results obtained in feature based steganalysis (80.3%) and neural network steganalysis (97.09%)
  9. Keywords:
  10. Deep Learning ; Convolutional Neural Network ; Steganography ; Joint Photo Graphic Expert (JPEG)Processor ; Steganalysis ; Binary Classifier

 Digital Object List

 Bookmark

No TOC