Active Steganalysis of MP3Stego

Abstract

The goal of steganalysis is to detect and/or estimate potentially hidden information from observed data. Steganalysis not only plays a significant role in information countermeasures, but also can prevent the illicit use of steganography. This paper develops an active steganalysis system for detecting hidden messages, by estimating frames with hidden message and message length in compressed audio files produced by MP3Stego. Principle Component Analysis (PCA) is applied not only to estimate uncorrelated components but also help to detect whether the received MP3 streams are steganography or original. In this steganalysis system, hidden messages can be detected in the PCA (whitening) stage. Independent Component Analysis (ICA) attempts to separate a set of steganographic signals from original signals. By analyzing the nature of MP3 signals (frame header, side information), frame with secret message can be detected. The results of empirical tests reach 97% and conclude that detection accuracy of the proposed steganalysis system is convenient for MP3Stego embedded contents. Experiments shows that the proposed method can be used quite effectively to detect locations of messages embedded using nature of MP3 signal.