Biometrics data recently has become a major role in determining the identity of the person. With such importance for the use of biometrics data, there are many attacks that threaten the security and integrity of biometrics data itself. Therefore, it becomes necessary to protect the originality of biometrics data against manipulation and fraud. This paper presents an authentication technique to achieve the authenticity of speech signals based on adaptive watermarking technique. The basic idea is depends on extracting the speech features from the speech signal initially and then using these features as a watermark. The watermark information embeds into the same speech signal. The short time energy technique is used to identifying the suitable positions for embedding the watermark in order to avoid the regions that used in the speech recognition system. After exclusion the important areas that used in speech recognition the Genetic Algorithm (GA) is used to generate random locations to hide the watermark information in an intelligent manner. The experimental results have achieved high efficiency in establishing the authenticity of speech signal and the process of embedding watermark not effect on features that used in speech recognition