Steganalysis Of Voice-over-IP Streams

The countermeasure technology of steganography is steganalysis whose goal is to detect, extract or break secret information hiding in suspicious carrier. Because of high steganographic capacity and instantaneity, steganography base on Vo IP present a range of security threats. For now, there still are some problems in steganalysis of Vo IP need to be solved. So in this paper, with the analysis of the problems which steganalysis of Vo IP facing, some researches has been done:1. Because of the different character between speech and image, most steganalysis method cannot be employed for speech. To solve the problem, we construct the parameter as gray degree of image by jointing random bits and bits for hiding information. The constructed parameter can be used for traditional steganalysis method on image like Chi-square. The experimental result shows that this method can get great accuracy on detection.2. We present a distributed steganalysis scheme for compressed speech in Voice-over-IP scenarios due to the experimental result shown that the detection performance is different with different parameters and features. In this scheme, each speech parameter available for concealing information is designed to be detected independently exploiting the corresponding optimal detection feature. At last, we evaluate their performance for steganalysis based on support-vector-machines with a large number of steganographic G.729 a speech samples at different embedding rates, and compare them with some existing algorithms. The result shows this scheme achieves higher performance than the existing algorithms.3. We present a new steganalysis scheme base on statistics character of parameter for detecting fixed codebook index. The method extracts variety of features by analyzing the characteristic of fixed codebook index and trains the classifier by support-vector-machine. The result shows that the scheme performance much better than the existing schemes in low embedding rate and short length speech. Even when the bit for embedding information is only one in a frame, this scheme can achieve high accuracy while the existing schemes could not detect the steganographic behavior.4. The structure of fixed codebook index on adaptive multi-rate codec is very special what makes the presented steganalysis scheme cannot performance well. For that, we design new feature for this codec and performance feature selection by Ada Boost for feature reduction. The selected features are trained by support-vector-machine and the trained classifier is employed to evaluate the performance by using different conditions with existing methods. The result shows that our scheme has obvious advantages in multiple indicators compare to the existing schemes.