In this paper, we propose a fragile watermark scheme for detecting the tampering or hostile attack of speech signal. Watermark block is generated by Gaussian Random operation on a null matrix, then the watermark is embedded into the original signal which is processed by DCT (discrete cosine transform). During the signal detection processing, matrix to vector method is adopted, in virtue of the differential matrix row and column, a threshold value can be calculated. With the comparison between the watermarked signal and threshold, the tampering location can be detected and the watermark scheme can achieve valid identification on tampering, with the accuracy rate reached up to 95%.
展开▼