AI-based Sensor Fingerprinting for Device Identification.

AI-based Sensor Fingerprinting for Device Identification.

Srivignessh Pss

Srivignessh Pss

Tempe, Arizona

Mobile/Wearable sensors have its own unique manufacturing defects. The AI developed learns these calibrations for identification of devices

Artificial Intelligence, Internet of Things

Description

Wearable and Mobile Device Accelerometer has linear calibration measurements. This provides us with two metrics sensitivity and offsets in each of the axes (X, Y, Z) to model. Hence there are 6 metrics for each device. Adaptive Neural Networks is used to learn these parameters to uniquely identify a device.

In the figure, You can see the POC. Three clusters represent (the Z axes) sensitivity and offsets of three devices. Example Devices used: Moto G, Mi4i, and Ipad.

References 1> https://crypto.stanford.edu/gyrophone/sensor_id.pdf 2> http://ieeexplore.ieee.org/document/7522434/

Links

Sensitivity vs Offset for Z axis

Standard index1

Medium img 20170629 212404

Edwin M. created project Farm Assist

Medium 905fb95f 6880 42b7 b6e5 7f583fb52138

Farm Assist

With everything going digital, there’s need for a digital platform that allows farmers to automate their farm activities such that they have visibility of the farm progress anytime anywhere on single click. Coming up with a digital online platform, Farm Assist (FA), an internet application software, will help farmers to solve these problems in a more efficient and less costly way leading to increase in production.

Medium 19657140 1605089572855250 3345190797171908989 n

Tushaar G. updated status

Medium 19657140 1605089572855250 3345190797171908989 n

Tushaar Gangarapu

Hello everyone, I am Tushaar. I am currently pursuing my bachelors at NITK. I developed passion towards AI and Machine Learning, had mentored the same at a mentorship program and currently was working on "Real Time Sensor based Weather Analysis and Prediction using Cloud Analytics to Improve Agricultural Yield". At the moment all we have are basic regression models which actually work with an 80 percent efficiency approximately. The repositories related to both the mentorship program and the regression models can be found on my GitHub account.

See More

No users to show at the moment.

No users to show at the moment.