Methods and systems for creating a synthetic speech detection algorithm and enhancing detection of fraudulent audio data using the algorithm
A method of creating an algorithm to enhance detection of fraudulent audio data is provided that includes generating, by an electronic device, audio data using a synthetic speech creation algorithm, processing the generated audio data using a synthetic speech detection algorithm, determining whether the synthetic speech detection algorithm detects the generated audio data, generating additional audio data using the synthetic speech creation algorithm, and creating a database including the generated audio data and genuine data. When the synthetic speech detection algorithm failed to detect the generated audio data, the synthetic speech detection algorithm is divided into a first algorithm and a second algorithm. The second algorithm is fine-tuned using the data in the created database. The first algorithm and the fine-tuned second algorithm are combined to create a fine-tuned synthetic speech detection algorithm for enhancing detection of fraudulent audio data generated by the synthetic speech creation algorithm.
1. A method of creating an algorithm to enhance detection of fraudulent audio data comprising the steps of:
generating, by an electronic device, audio data using a synthetic speech creation algorithm;
processing the generated audio data using a synthetic speech detection algorithm;
determining whether the synthetic speech detection algorithm detects the generated audio data;
in response to determining the synthetic speech detection algorithm failed to detect the generated audio data, generating additional audio data using the synthetic speech creation algorithm and creating a database including the generated audio data and genuine data;
dividing the synthetic speech detection algorithm into a first algorithm and a second algorithm;
fine-tuning the second algorithm using the generated audio data and genuine data in the created database, the fine-tuned second algorithm being different than the second algorithm; and
combining the first algorithm and the fine-tuned second algorithm to create a fine-tuned synthetic speech detection algorithm for enhancing detection of fraudulent audio data generated by the synthetic speech creation algorithm.
2. The method according to claim 1 , further comprising determining the synthetic speech detection algorithm is capable of detecting fraudulent audio data using audio data generated by the synthetic speech creation algorithm, when the synthetic speech detection algorithm detects the generated audio data.
3. The method according to claim 1 , further comprising the step of creating the synthetic speech detection algorithm.
4. The method according to claim 3 , wherein the synthetic speech detection algorithm is a voice cloning algorithm.
5. An electronic device for creating an algorithm to enhance detection of fraudulent audio data comprising:
a processor; and
a memory configured to store data, said electronic device being associated with a network and said memory being in communication with said processor and having instructions stored thereon which, when read and executed by said processor, cause said electronic device to:
generate audio data using a synthetic speech creation algorithm;
process the generated audio data using a synthetic speech detection algorithm;
determine whether the synthetic speech detection algorithm detects the generated audio data;
in response to determining the synthetic speech detection algorithm failed to detect the generated audio data generate additional audio data using the synthetic speech creation algorithm; and create a database including the generated audio data and genuine data;
divide the synthetic speech detection algorithm into a first algorithm and a second algorithm;
fine-tune the second algorithm using the generated audio data and genuine data in the created database, the fine-tuned second algorithm being different than the second algorithm; and
combining the first algorithm and the fine-tuned second algorithm to create a fine-tuned synthetic speech detection algorithm for enhancing detection of fraudulent audio data generated by the synthetic speech creation algorithm.
6. The electronic device according to claim 5 , wherein the instructions when read and executed by said processor, cause said electronic device to determine the detection algorithm is capable of detecting fraudulent audio data using audio data generated by the synthetic speech creation algorithm when the detection algorithm detects the generated audio data.
7. The electronic device according to claim 5 , wherein the instructions when read and executed by said processor, cause said electronic device to create the synthetic speech detection algorithm.
8. The electronic device according to claim 7 , wherein the synthetic speech creation algorithm is a voice cloning algorithm.
9. A non-transitory computer-readable recording medium in an electronic device for training an algorithm to enhance detection of fraudulent audio data, the non-transitory computer-readable recording medium storing instructions which when executed by a hardware processor cause the non-transitory recording medium to perform steps comprising:
generating audio data using a synthetic speech creation algorithm;
processing the generated audio data using a synthetic speech detection algorithm;
determining whether the synthetic speech detection algorithm detects the generated audio data;
in response to determining the detection algorithm failed to detect the generated audio data, generating additional audio data using the synthetic speech creation algorithm and; creating a database including the generated audio data and genuine data;
dividing the detection algorithm into a first algorithm and a second algorithm, wherein the first algorithm remains unchanged and the second algorithm is changeable;
fine-tuning the second algorithm using the generated audio data and genuine data in the created database, the fine-tuned second algorithm being different than the second algorithm; and
combining the first algorithm and the fine-tuned second algorithm to create a fine-tuned synthetic speech detection algorithm for enhancing detection of fraudulent audio data generated by the synthetic speech creation algorithm.
10. The non-transitory computer-readable recording medium according to claim 9 , wherein the instructions when read and executed by said processor, cause said non-transitory computer-readable recording medium to perform the step of determining the synthetic speech detection algorithm is capable of detecting fraudulent audio data using audio data generated by the synthetic speech creation algorithm, when the synthetic speech detection algorithm detects the generated audio data.
11. The non-transitory computer-readable recording medium according to claim 9 , wherein the instructions when read and executed by said processor, cause said non-transitory computer-readable recording medium to perform the step of creating the synthetic speech detection algorithm.
12. The non-transitory computer-readable recording medium according to claim 11 , wherein the synthetic speech creation algorithm is a voice cloning algorithm.