Research on a New Speech Synthesis Fraud Method

With the rapid development of deep learning technology, deep counterfeiting has also quietly ascended to the stage of history, and voice disguise fraud, as a new type of attack means, has laid a huge hidden danger for people's privacy security. Based on the research of speech synthesis, this paper designs and implements a speech disguise synthesis system based on the Tacotron2 model, and uses the WaveGlow model as a vocoder to replace the original WaveNet module, which is used to simulate experimental research from the direction of the attack to achieve a better synthesis effect. After this experiment, this paper verifies the feasibility of the program and provides technical references for subsequent security technology development.