Abstract

Mp3 is a very popular audio format and hence it can be a good host for carrying hidden messages. Therefore, different steganography methods have been proposed for mp3 hosts. UnderMp3Cover is one of such algorithms and has some important advantage over other comparable methods. First, the popular steganography method mp3stego, works directly on non-compressed samples. Therefore, using covers that have been compressed before could lead to serious degradation of its security. UnderMp3Cover does not have this important limitation. Second, cover of UnderMp3Cover may be generated from different mp3 encoders. This feature could improve security of UnderMp3Cover. In order to make UnderMp3Cover accessible to other researchers, this steganography method was simulated in Matlab environment and also some of its security problems were addressed. Also, an executable file with its corresponding Matlab wrapper were written to facilitate the decoding process of mp3 frames and importing information of mp3 files into Matlab environment. Finally, Matlab codes for extracting multiple re-embedding calibrated features for steganalysis of UnderMp3Cover are provided. These steganalysis features are based on estimating the mutual information between global gain and other fields of side information.

Instructions:

The archive file “Mp3_Decoder” contains some codes for decoding and reading frames of mp3 files into Matlab. You need to decompress the archive before using the codes.

Upon decompression, the resulting folder would contain an executable file, a folder named “tables” and the Matlab wrapper for communication between Matlab environment

and the executable part. Please, make sure that you copy all of these material to your current address and do not make any change to the name of “tables” folder.

Please, refer to “Main_Frame_Decoder.m” for an example on running the codes.

The folder “Multiple_ReEmbeding_Feature_Extraction” contains the codes for proposed feature extraction. First, you should use the codes in “Mp3_Decoder” folder and

import sideinfo of your mp3 files in Matlab, then use the provided codes for feature extraction.Please, refer to “Main_Feat_Extraction.m” for an example on running the codes.

The archive file “Ump3C” contains some codes for message hiding/extraction in the Matlab environment. Investigating Ump3C method shows that only global gain of mp3

bit stream get changed in the embedding process, therefore, our code is only a simulation of the process and does not produce an mp3 file in the output.

In other words, these codes would only generate sideinfo corresponding to the stego files, but it would not mux sideinfo and MDCT back, it would not produce

bit stream of stego file and it would not write the bitstream into an output file. It is noteworthy that from steganalysis point of view, we don’t need the output

bit stream of Ump3c and sideinfo of stego file would suffice. In order to run the codes, you need to decompress the archive. Upon decompression, the resulting folder

would contain an executable file, a folder named “tables” and some Matlab codes for simulating the message hiding/extraction procedures. Please, make sure that you

copy all of these material to your current address and do not make any change to the name of “tables” folder. Please, refer to “Main.m” for an example on running the codes.

Fore more information please refer to the following article:

Hamzeh Ghasemzade, "Multi-layer architecture for efficient steganalysis of UnderMp3Cover in multi-encoder scenario", IEEE Transactions on Information Forensics and Security, 2018.