Information Hiding Algorithm Of Speech Signal Based On Multi-Statistics Analysis

Over the past ten years, among various information security techniques, a new sub-discipline called information hiding has drawn tremendous attention of academe and correlative area. As a new branch of information security, information hiding technology including digital watermarking and digital steganography aims to protect intellectual property rights of multimedia contents and to send secret messages under the cover of a carrier signal. Digital steganography is used to exchange secret messages by military units or intelligence agencies, or protect private or sensitive data by companies or individuals. But recently this technology is abused to steal the secret of companies by staffs; the terrorist organizations also utilize digital steganography to hide their communications from law enforcement. So the research on this domain is the critical to prevent steganography from being abused, and is significant to security and social stabilization.This thesis studies digital steganography which uses speech signal as carriers. A robust information hiding algorithm which embeds secret information in wavelet domain of speech signal was designed in this thesis for better solidity and security goals after surveying the appropriate background literature. At last we discuss some problems which await us to solve, and more aspects of research are presented. The main contributions of the thesis are as follows:This paper describes a new information hiding algorithm in wavelet domain of speech signal based on multi- statistics analysis. At First, voiced segments are detected from original speech and then transformed by wavelet decomposition; low frequency coefficients of the transformation are withdrawn. Then the coefficients of each frames are separated into two groups, the energy and variance of absolute coefficients value are computed for each group. According to statistics comparison and secret information bits, we hide information by not changing or slightly adjusting the value of statisticsThis algorithm only hides information in voiced speech, so it fully utilizes human auditory characteristic and has a good security. It improves solidity of algorithm by hiding information in multi- statistics. Experiment results show that secret information can be extracted blindly. The proposed algorithm is robust to attack of noise, low-pass filtering, re-sampling and re-quantization.