Researchers from the Language Technologies Institute (LTI) of Carnegie Mellon University have developed a way to protect voice-recognition identification systems from fraud and identity theft.
The system developed by computer scientist and lead researcher Bhiksha Raj traces the multiple ways in which a person speaks and converts them into hundreds of alphanumeric sequences using different mathematical functions. By comparing how many of those match, it can determine whether the speaker is the person who enrolled. To make it even more secure, the system throws in an extra data specific to your device, so that nobody else besides the owner can generate the specific strings that he or she did.
In the first tests that were performed, the accuracy rate was of 95%. Even though it’s not as good as other system, improvements are being made to reach a 100% of success.