METHOD FOR MULTIFACTOR AUTHENTICATION USING BONE CONDUCTION AND AUDIO SIGNALS
Provided here are systems and method for two-way authentication of a user. In an embodiment, the method may include, in response to reception of an audio signal from a user, determining, via a smart device and/or authentication circuitry, whether a corresponding bone conduction signal is received from one of one or more separate wearable devices. The method may include, in response to determination that the corresponding bone conduction signal is received: determining, via the smart device and/or authentication circuitry, whether the audio signal and corresponding conduction signal are consistent. The method may include, in response to a determination that the audio signal and corresponding conduction signal are consistent: verifying, via the smart device, the audio signal; verifying, via the smart device, the corresponding bone conduction signal; and, in response to verification of the audio signal and the corresponding bone conduction signal, authenticating a user.
1 . A method for two-way authentication of a user, the method comprising:
receiving a bone conduction signal from the user via one or more wearable devices;
receiving an audio signal from the user via a microphone separate from the one or more wearable devices, the audio signal to correspond with the bone conduction signal;
determining a consistency score for the audio signal and corresponding bone conduction signal;
in response to the consistency score being greater than or equal to a consistency threshold:
verifying, using an audio conduction model (AC model), that the audio signal is associated with the user;
verifying, using a bone conduction model (BC model), that the bone conduction signal is associated with the user; and
in response to verification of the audio signal and the bone conduction signal, enabling, for the user, access to a smart device.
2 . The method of claim 1 , wherein determination of the consistency score includes:
determining a marginal bone conduction power distribution with respect to time;
selecting a time range of interest based on an average of the margin marginal bone conduction power distribution with respect to time;
determining a marginal audio conduction and bone conduction power distribution with respect to frequency;
selecting a top frequency index M from the marginal bone conduction power distribution with respect to frequency and a top frequency index N from the marginal audio conduction power distribution with respect to frequency;
determining a correlation matrix between the top frequency index M and the top frequency index N; and
generating, based on the correlation matrix, the consistency score.
3 . The method of claim 1 , further comprising, prior to determining the consistency score, pre-processing the bone conduction signal.
4 . The method of claim 3 , wherein pre-processing the bone conduction signal includes passing the bone conduction signal through (1) a low-pass filter to remove noise generated by human motion and (2) a Wiener filter to remove the noise.
5 . The method of claim 1 , further comprising, prior to determining the consistency score, passing the audio signal through a Wiener filter to remove noise.
6 . The method of claim 1 , further comprising, prior to determination of a consistency score and verification, prompting the user to submit an initial or enrollment (1) bone conduction signal and (2) audio signal.
7 . The method of claim 6 , further comprising:
training the AC model with a plurality of responses prior to submission of an initial or enrollment audio signal, and
training the AC model with a submitted initial or enrollment audio signal.
8 . The method of claim 6 , wherein the BC model comprises a convolutional neural network (CNN).
9 . The method of claim 8 , further comprising training the CNN with a stored/pre-collected bone conduction dataset.
10 . The method of claim 1 , further comprising, if one or more of the audio signal or bone conduction signal are not verified, preventing the user from accessing the smart device.
11 . The method of claim 1 , further comprising, prior to determining the consistency score, delaying an earliest received of the bone conduction signal and the audio signal to align the bone conduction signal and the audio signal.
12 . A method for two-way authentication of a user, the method comprising:
prompting the user to submit initial or enrollment (a) audio signals and (b) bone conduction signals;
updating an audio conduction model (AC model) and a bone conduction model (BC model) based on the received initial or enrollment (a) audio signals and (b) bone conduction signals;
after reception of the initial or enrollment (a) audio signals and (b) bone conduction signals:
receiving a bone conduction signal from the user via one or more wearable devices;
receiving an audio signal from the user via a microphone separate from the one or more wearable devices, the audio signal corresponding to the bone conduction signal;
determining a consistency score for the audio signal and corresponding bone conduction signal;
in response to the consistency score being greater than or equal to a consistency threshold:
verifying, using the AC model, that the audio signal is associated with the user;
verifying, using the BC model, that the bone conduction signal is associated with the user; and
in response to verification of the audio signal and bone conduction signal, enabling, for the user, access to a smart device.
13 . The method of claim 12 , wherein the initial or enrollment audio signal is received via the microphone, and wherein the smart device includes the microphone.
14 . The method of claim 13 , wherein the initial or enrollment bone conduction signals are received from one of one or more wearable devices.
15 . The method of claim 12 , further comprising prompting the user to submit initial or enrollment (a) audio signals and (b) bone conduction signals for each of the one or more wearable devices.
16 . The method of claim 15 , wherein the initial or enrollment (a) audio signals and (b) bone conduction signals include one or more specific phrases.
17 . The method of claim 12 , wherein the BC model comprises a convolutional neural network (CNN), and wherein the CNN is trained using a stored/pre-collected bone conduction dataset to generate corresponding initial embedded bone conduction vectors.
18 . The method of claim 17 , further comprising, prior to verification via the BC model, generating embedded bone conduction vectors using the bone conduction signals and the initial embedded bone conduction vectors.
19 . A non-transitory machine-readable storage medium storing processor-executable instructions that, when executed by at least one processor, cause the at least one processor to:
receive a bone conduction signal from a user via one or more wearable devices;
receive an audio signal from the user via a microphone separate from the one or more wearable devices, the audio signal corresponding to the bone conduction signal;
determine a consistency score for the audio signal and the bone conduction signal;
in response to the consistency score being greater than or equal to a consistency threshold:
verify, using an audio conduction model (AC model), that the audio signal is associated with the user;
verify, using a bone conduction model (BC model), that the bone conduction signal is associated with the user; and
in response to verification of the audio signal and bone conduction signal, enable, for the user, access to a smart device.
20 . The non-transitory machine-readable storage medium of claim 19 , wherein the smart device includes the microphone.
21 . The non-transitory machine-readable storage medium of claim 20 , wherein the bone conduction signal is received via wireless communication.
22 . The non-transitory machine-readable storage medium of claim 19 , wherein the consistency threshold is based on similarities between bone conduction signals and audio signals that indicate the bone conduction signal and the audio signal are from the user.