Loading...

Steganalysis of Audio Signals Based on Discriminative Features Statistics

Haji Shir Mohammadi, Mahmood | 2013

594 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 43979 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Gholampour, Iman; Ghaemmaghami, Shahrokh
  7. Abstract:
  8. Digital Steganography is the embedding of digital data into a video, image or audio signal. The embedding must be done such that the perceptual quality of the host signal does not seriously degrade. In the last decade, Steganography has received considerable attention in various application fields. Steganography is not always a legal process and may be used to establish illegal communications, transfer malware files to certain targets, or collect data illicitly from an organization. Steganalysis methods are developed to detect Steganography in such applications.
    The purpose of this thesis is to develop a new steganalysis method based on perceptual models and statistical structure of the signal that uses the spectro-temporal features of audio signals. These features consist of two categories. One of them contains neighborhood characteristics of audio samples in the time domain, and the other is based on the higher-order statistics of the signal. Using these two sets of features together and a recursive elimination algorithm, features with the greatest impact are selected. Machine learning will help to examine the accuracy of the proposed method.
    To evaluate the performance of the proposed method, it is applied to MP3 and WAV audio formats which are embedded with MP3Stego and Invisible Secret, and then we check the performance of the new method by comparing it to some well-known steganalysis methods. For the MP3 format, accuracy of 84.1% is achieved and the WAV format steganalysis accuracy is about 78.5%. To further study, we apply the proposed method to the MP3 files which are embedded using Perturbed Quantization method, where the detection accuracy of 76.57% is achieved
  9. Keywords:
  10. Steganography ; Learning Machine ; Audio Features ; Higher Order Statisyics ; Steganalysis

 Digital Object List

 Bookmark

No TOC