The development of smartphones increasingly affects the human life, such as data security. The use of smartphones to maintain data security is still low, which generally only uses conventional method as security systems available on smartphone devices, such as word combinations, PINs or patterns. Various weaknesses of conventional methods cause the development of biometrics systems. Therefore, a technology is created to replace the conventional security system that is biometric security system using natural human characteristics, one of them using voice. Data security uses the android based sound biometric app using CMUSphinx as the word library. CMUSphinx does not require an internet connection to run it. MFCC (Mel Frequency Cepstrum Coefficients) is as a characteristic extraction method on a digital sound signal. Matching process with saved data uses DTW (Dynamic Time Warping) method. This study can help improve the data security of smartphone; therefore, it cannot be sabotaged or accessed by other parties that are not desired.