Back to results
US 11,695,763 B2Granted
System and method for fingerprinting a device based on sensor data
AssigneePaypal, Inc.
InventorsZhe Chen, Hewen Wang
Life of this patent
Estimating the term…
Abstract
Methods and systems are presented for generating a device fingerprint based on data obtained from one or more sensors on a device. A plurality of data points corresponding to sensor readings are obtained from the one or more sensors on the device. A set of time-domain features and a set of frequency-domain features are extracted from the plurality of data points and inputted to a neural network trained using a triplet network. A device fingerprint that may be used to identify the device is obtained from the neural network.
Claims20
A system, comprising:
- a non-transitory memory;
- and
- one or more hardware processors coupled with the non-transitory memory and configured to read instructions from the non-transitory memory to cause the system to perform operations comprising:
- obtaining, from a sensor on a first device, a plurality of data points based on a first set of readings from the sensor, wherein each data point of the plurality of data points is a tuple including a number of elements;
- extracting a set of time-domain features from the plurality of data points, wherein the extracting comprises:
- determining a series of root sum squares, wherein each item in the series corresponds to a root sum square of the elements of the tuple for each data point of the plurality of data points, and
- obtaining a resampled plurality of data points by performing a cubic spline interpolation on the series of root sum squares, wherein the set of time-domain features are extracted from the resampled plurality of data points;
- extracting a set of frequency-domain features from the plurality of data points;
- inputting feature data for the set of time-domain features and the set of frequency-domain features into an input layer for neural network features of a neural network;
- processing the feature data from the input layer using a plurality of hidden layers of the neural network;
- computing a vector for the neural network features from the set of time-domain features and the set of frequency-domain features using one of the plurality of hidden layers, wherein the vector provides an increased security for the first device during device fingerprinting;
- and
- obtaining, from an output layer of the neural network, a first device fingerprint, wherein the first device fingerprint comprises the vector as a representation of the neural network features for the feature data.
10 more elements
A method, comprising:
- receiving, at a web server, a request from a first device associated with a first device ID;
- obtaining, from a first sensor on the first device, a first plurality of data points based on readings from the first sensor;
- inputting feature data for the first plurality of data points into an input layer for neural network features of a neural network;
- processing the feature data from the input layer using a plurality of hidden layers of the neural network;
- computing a vector for the neural network features from the first plurality of data points using one of the plurality of hidden layers, wherein the vector provides an increased security for the first device during device fingerprinting;
- obtaining, from a neural network, a first device fingerprint based on the first plurality of data points, wherein the first device fingerprint comprises the vector as a representation of the neural network features for the feature data;
- retrieving, from a data store, a second device fingerprint associated with the first device ID;
- determining a distance between the first device fingerprint and the second device fingerprint;
- retrieving, from a data store, a plurality of sets of device fingerprints, each set of device fingerprints being associated with a device ID;
- determining a best-matching set of device fingerprints based on comparisons between the first device fingerprint and the each set of device fingerprints;
- and
- determining the device ID associated with the best-matching set of device fingerprints corresponds to the first device based at least on the device ID being associated with the best-matching set of device fingerprints.
9 more elements
A non-transitory machine-readable medium having stored thereon machine-readable instructions executable to cause a machine to perform operations comprising:
- obtaining, from a first sensor on a first device, a first plurality of data points based on readings from the first sensor;
- obtaining, from a second sensor on a first device, a second plurality of data points based on readings from the second sensor;
- extracting a first set of time-domain features from the first plurality of data points and a second set of time-domain features from the second plurality of data points;
- extracting a first set of frequency-domain features from the first plurality of data points and a second set of frequency-domain features from the second plurality of data points, wherein the extracting of the first and second sets of frequency-domain features comprises:
- obtaining a frequency-domain representation of a resampled plurality of data points by performing a discrete Fourier transform on the resampled plurality of data points, wherein the resampled plurality of data points are associated with elements for the first and second set of time-domain features, and wherein the first and second set of frequency-domain features are extracted from the frequency-domain representation of the resampled plurality of data points;
- inputting feature data for the first set of time-domain features and the second set of time-domain features and the first set of frequency-domain features and second set of frequency-domain features into an input layer for neural network features of a neural network;
- processing the feature data from the input layer using a plurality of hidden layers of the neural network;
- computing a vector for the neural network features from the set of time-domain features and the set of frequency-domain features using one of the plurality of hidden layers, wherein the vector provides an increased security for the first device during device fingerprinting;
- and
- obtaining, from an output layer of the neural network, a first device fingerprint wherein the first device fingerprint comprises the vector as a representation of the neural network features for the feature data.
7 more elements
Dependencies are inferred from the claim text.