A lot of science fiction tells us that one day, man will interact with machine by talking to it, and that machine will be able to recognize who is talking. Already we have this to a degree, as voice prints can be used as security and new tools like Siri are developed, but there are some issues with these systems that researchers at Carnegie Mellon University hope to address.
To unlock the full power of Siri and other voice assistants, the audio of your voice has to be sent to a server for processing. This is understandable as a smartphone does not have the power to do that processing itself, but this transmission can be dangerous. An eavesdropper could catch and copy the audio information, and then use that to access anything that requires your voice print. What the researchers came up with though is a way to actually represent your voice print as strings of data. These strings can then be used like a typical password, and by also including another string that uniquely identifies the device, the system can be more secure than current voice-print security.
When tested, the system demonstrated 95% accuracy at recognizing the user's voice print, which is quite promising for this to be implemented. Before we see this used in actual products though, we may first be forced to consider the privacy issues of other speech-based services. After all, your voice carries with it a lot of identifying information.