Steganalysis for MP3Stego using calibrated side information features
XianminYu Rangding Wang Diqun Yan
School of Information Science and Engineering, Ningbo University, Ningbo 315211, China
e-mail:wangrangding@nbu.edu.com
Abstract: MP3Stego is a typical steganographic
tool for MP3 audio, which embeds secret message
into MP3 audio according to the parity of the block
length. In this work, we present a steganalytic
method to attack MP3Stego. The big_values in side
information is considered to extract the
steganalytic feature. Re-compression calibration
has been applied in order to improve the feature's
sensitivity. Experimental results show that the
extracted feature can reflect the MP3Stego trace
effectively.
Keywords-steganalysis;MP3; big_values; calibration
I. INTRODUCTION
Steganography is used to covey secret
message by concealing their exsitence into an
innocent carrier[1]. The countermeasure to
steganography is steganalysis, and its aim is to
reveal the presence of the secret message. Due to
the suitability of MP3 audio as steganographic
cover[2], several stego tools for MP3 audio have
been arisen such as MP3Stego[3],
UnderMP3Cover[4], Stego-Lame[5]. Among
them, MP3Stego is a most typical one. It can
achieve so good imperceptibility that it is hard to
distinguish between background noise and
steganographic distortion when the original
audio is unavailable.
To attack MP3Stego, Westfeld [6] proposed a
steganalystic method based on the variance of
the block lengths. Qiao[7] introduced a detection
method based on an inter-frame feature set
which contains the moment statistical features
on the second derivatives, as well as Markov
transition features and neighboring joint density
of the MDCT coefficients. In his o ther work[8],
the statistical moments of GGD (Generalized
Gaussian Distribution) shape parameters f or
MDCT coefficients are also extracted as the
steganalytic features. Ozer [9] presented a
method for detecting audio steganography based
on audio quality metrics. Since MP3Stego
happens during quantization, more attentions
should be paid to the parameters associated with
the quantization. In this work, the big_values in
side information which is closely related to
quantization is introduced into steganalysis to
attack MP3Stego. In order to weaken the
influence of the audio content, the calibrated
feature is extracted by re-compression.
The rest of this paper is organized as follows.
Section 2 briefly covers the basic operations of
the MP3Stego algorithm. Section 3 presents the
proposed method. The effectiveness of the
proposed method under various conditions with
experimental results is verified in Section 4.
Finally, conclusions are drawn in Section 5.
II. R
EVIEW OF MP3STEGO ALGORITHM
MP3Stego embeds compressed and
encrypted data in the inner loop which is used
for quantization and coding during MP3 coding.
The inner loop quantizes the MDCT
coefficients and increases the quantization step
until the quantized coefficients can be coded
with the available number of bits. The variable
part2_3length contains the number of the bits
used for scalefactors and Huffman code data for
current granule. Without embedding, the inner
loop will be finished when the part2_3length is
within the range of the number of bits available.
In MP3Stego, the inner loop will continue to
iterate until the parity of the part2_3length is
equal to the hidden bit and the bit demand for
Huffman coding is also met. Once the inner loop
is done, another loop, namely outer loop, will
530
978-0-7695-4806-7/12/$26.00 ©2012 IEEE