Ruby Lee and Guangyuan Hu: Tiny AI module for detecting smartphone theft and anomalous behavior

Nov. 2, 2021
Ruby Lee

A built-in hardware system could rapidly detect when a thief tries to use a stolen cell phone to access the phone’s data and online information.

The system uses artificial intelligence (AI) to evaluate data from the phone’s motion sensors to distinguish if the phone is being used by its rightful owner. Phone thieves move and manipulate the phone in a different way than would its true owner. Data collected from the phone’s motion sensors, such as its accelerometer and gyroscope, can yield a profile of the rightful owner.

The new system, known as Smartphone Imposter Detector, uses an AI technique known as deep learning enhanced with statistical tests to quickly detect anomalous handling of the phone. The system can then prevent access to sensitive information or completely shut the phone down. The smartphone owner could set preferences so that only trusted users are allowed.

The approach is supported by a tiny hardware module that can be added to the phone to implement the Smartphone Imposter Detector algorithm. The module adds minimal energy consumption and outperforms existing machine-learning algorithms. Through the use of deep learning, the module learns the phone owner’s behavior patterns. Unlike other approaches, the module never shares this information via the cloud, significantly reducing user data exposure.

The module could be used beyond the smartphone application to detect anomalous behavior in critical infrastructures like the power grid, other cyber-physical systems and internet-of-thing (IOT) devices. It can detect anomalous behaviors by using hardware event counters built into microprocessors rather than motion sensors.

Guangyuan Hu

"We have designed a hardware AI module that is very simple to implement, does not require any exotic technologies, and yet can defeat attackers, from smartphone imposters to power-grid attackers.”
– Ruby Lee

Ruby Lee, Forrest G. Hamrick Professor
in Engineering and Professor of Electrical
and Computer Engineering

Guangyuan Hu, Graduate Student
in Electrical and Computer Engineering

Team members:
Zecheng He, Graduate Student
in Electrical and Computer Engineering

Development status:
Patent protection is pending. The technology is
nonexclusively licensed and Princeton is continuing
to seek outside interest for the development
of this technology.

A person stealing a phone from a bag

A tiny AI module can detect cell-phone theft and be adapted to monitor cyber-security systems and other smart devices.

National Science Foundation
Semiconductor Research Corporation

Learn more:
Email: [email protected]

Licensing contact:
Chris Wright
Licensing Associate
Office of Technology Licensing
Email: [email protected]