Elsevier

Signal Processing

Volume 91, Issue 4, April 2011, Pages 877-889
Signal Processing

An active steganalysis approach for echo hiding based on Sliding Windowed Cepstrum

https://doi.org/10.1016/j.sigpro.2010.09.006Get rights and content

Abstract

This paper presents an active steganalysis technique for echo hiding (EH). This approach can differentiate stego audios (with hidden message) from nature ones (without hidden message), and then extract sequential EH messages in stego audios without prior knowledge about the EH technique. The procedure of this algorithm can be simplified as follows: (i) Sliding Windowed Cepstrum (SWC) is obtained by employing a Sliding Window (SW) to move over the audio signal sample by sample. (ii) The echo detector is designed based on statistical analysis of the Cepstrum Peak Location Aggregation Degree (CPLAD). (iii) For stego signals, both echo delay and the range of segment length are estimated by exploring the changing rule of the SWC. (iv) The accurate estimation of segment length and synchronism position is acquired by the grid search algorithm. (v) Finally, the binary secret message can be extracted using the estimated parameters in the previous steps. The proposed method has been implemented and tested to extract the message which is embedded into audio clips by using EH techniques with different kernels. The experimental results show that the accuracy of extraction is no less than 90% when attenuation is greater than 0.45.

Introduction

Steganography is the art and science of hiding the very presence of communication by embedding secret messages into cover signal, such as digital image, video, and audio [1], [2], [3], [4], [5], [6]. To achieve covert communication, stego signal should be indistinguishable from cover signal. Steganalysis is to detect and/or estimate potential secret messages in a media with little or no prior knowledge of steganography algorithms. Generally, steganalysis is classified into two categories [7]: passive steganalysis [8], [9] and active steganalysis [10], [11], [12], [13]. The former focuses on detecting the existence of secret messages, while the latter focuses on estimating some parameters, and even extracting the secret messages.

Many audio steganography techniques have been introduced during the past few years, such as technique based on masking [14], spread spectrum [3], [15], phase coding [16], EH [5], [6], [17], [18], [19], [20], [21], etc. EH is a method of hiding information in an audio clip by the addition of imperceptible echoes. It has been widely used because the encoding and decoding process is simple and EH is robust and imperceptible. Typical kernel (denoted as T-kernel) EH system was first introduced by Bender et al. [5] and Gruhl et al. [6]. Oh et al. presented the positive-and-negative kernels (denoted as PN-kernels) EH scheme [17]. Kim and Choi proposed EH scheme with backward-and-forward kernels (denoted as BF-kernels) [18]. Several modifications have been developed to improve the robustness and capacity [19], [20], [21].

Craver et al. [22] applied EH as general multiplicative embedding in the frequency domain and derived appropriate detectors based on various statistical models of audio FFT coefficients. Chen and Zhu [23] introduced a novel echo detection scheme employing autocorrelation of power cepstrum. Zeng et al. [24], [25] implemented an effective steganalysis method by analyzing the statistical moments of peak frequency which is obtained by short window extraction.

The previously discussed echo steganalysis techniques are passive steganaysis. The available active steganalysis techniques are almost suitable for additive embedding methods, such as Direct Sequence Spread Spectrum (DSSS) [10], [11], [12], Least Bit Plane (LSB) [13], etc. To our knowledge, there is hardly any literature discussing active steganalysis techniques for the EH method, which employs convolution embedding. In this paper, we propose an active steganalysis approach for EH based on Sliding Windowed Cepstrum (SWC). It cannot only detect the existence of secret message in the suspicious carriers, but also estimate embedding parameters (such as echo delay, segment length, and synchronism position) and extract sequential EH messages in stego audios. First of all, an echo detector is designed by statistical analysis of the Cepstrum Peak Location Aggregation Rate (CPLAR). Then, the echo delay and the range of segment length are estimated for stego audio; next, the accurate segment length and synchronism position are obtained by using the grid search technique; at last, the binary secret message is extracted with the estimated parameters in the previous steps.

This paper is organized as follows. Section 2 presents the scope of the EH method. The echo detector design and delay estimation by SWC are introduced in Section 3. Section 4 discusses the details of segment length and synchronism position estimation. Experimental results are presented in Section 5, and Section 6 draws the conclusions.

Section snippets

Echo hiding

EH scheme was first introduced by Bender et al. [5] and Gruhl et al. [6]. The echo kernel is expressed ash(n)=δ(n)+αδ(nd).

The delta function αδ(nd) adds the echo signal to the cover signal with a delay offset d and echo attenuation α. Let x(n) represent the cover audio, the stego signal y(n) can be expressed by the convolution between x(n) and h(n):y(n)=h(n)x(n).

The encoding procedure can be described as follows:

  • (1)

    Divide the cover audio x(n) into consecutive segments with same length.

  • (2)

    Embed one

Echo detector and delay estimation by SWC

In contrast with the decoding algorithm, steganalyst cannot get any information of steganography algorithm, such as segment length, synchronous position, delay offset, etc. Consequently, it is impossible to directly use cepstrum to steganalyze stego signal, but SWC can help.

Segment length and synchronism position estimation

Besides delay offset, segment length and synchronism position are another two key parameters for the EH method. In the proposed method, the estimation for these two parameters is implemented by two steps: rough estimation and accurate estimation.

Experimental results

The proposed steganalysis scheme has been tested on an audio set including 200 nature audio clips and 600 stego clips, which cover speech, music, and songs etc. These clips are sampled at 22.05 or 44.1 kHz with 16 bits per sample.

The stego audio clips are generated by three types of echo kernel steganography: T-kernel, PN-kernels, and BF-kernels. Text messages are embedded into these clips. According to the embedding parameters (segment length N, delay d1 and d0), all the 600 stego audios are

Conclusions

In this paper, we presented an active steganalysis technique for EH method based on exploring the changing rule of SWC as the window sliding over the audio signal. The proposed algorithm first detect the existence of secret message in the suspicious carriers, and then estimate the important embedding parameters, including echo delay, segment length, and synchronism position for the stego files, the sequentially EH messages is extracted finally. The method has been implemented and tested to

References (31)

  • K. Wong et al.

    A DCT-based mod4 steganographic method

    Signal Processing

    (2007)
  • M. Wu et al.

    Multimedia Data Hiding

    (2003)
  • I.J. Cox et al.

    Secure spread spectrum watermarking for multimedia

    IEEE Transactions on Image Processing

    (December 1997)
  • W. Bender et al.

    Techniques for data hiding

    IBM System Journal

    (1996)
  • D. Gruhl, A. Lu, W. Bender, Echo hiding, in: Proceedings of the First International Workshop on Information Hiding,...
  • R. Chandramouli

    A mathematical framework for active steganalysis

    Multimedia Systems

    (2003)
  • Sorina Dumitrescu et al.

    Detection of LSB steganography via sample pair analysis

    IEEE Transactions on Signal Processing

    (July 2003)
  • Ismail Avcıbas et al.

    Steganalysis using image quality metrics

    IEEE Transactions on Image Processing

    (2003)
  • R. Chandramouli, K.P. Subbalakshmi, Active steganalysis of spread spectrum image steganography, in: Proceedings of the...
  • S. Sedghi, H.R. Mashhadi, M. Khademi, Detecting hidden information from a spread spectrum watermarked signal by genetic...
  • F. Fan, J. Wang, X. Liu, H. Feng, An active steganalysis method of block-DCT image information hiding, electronic...
  • B. Xu, Z. Zhang, J. Wang, X. Liu, Improved BSS based schemes for active steganalysis, software engineering, artificial...
  • K. Gopalan, A unified audio and image steganography by spectrum modification, industrial technology, ICIT 2009, in:...
  • H. Matsuoka, Spread spectrum audio steganography using sub-band phase shifting, intelligent information hiding and...
  • Cited by (9)

    View all citing articles on Scopus
    View full text