Perception in a Mobile Device | Lucideus Research
Introduction:
“Perception (from the Latin perceptio) is the organisation, identification, and interpretation of sensory information in order to represent and understand the presented information, or the environment.” Living beings residing in the natural environment use their senses like sight, taste, touch, smell and hearing to perceive and comprehend different features of their surroundings. This sensory data is used by the brain at a subconscious level to create a model of the behavior pattern of things in the environment, which is then used in the interactions. This is exactly what a pet animal does as it gets introduced to a new owner and with time it gets comfortable around that person.
Imagine teaching perception to a mobile device where we consider the mobile device to be like a pet for its owner. The sensory organs in this case would be the sensors present on a device like accelerometer, proximity sensor, gyroscope, light sensor, etc. This sensor data is continuously collected by a device and can be used to build a dynamically evolving behavior profile for the primary user of the device.
“Perception (from the Latin perceptio) is the organisation, identification, and interpretation of sensory information in order to represent and understand the presented information, or the environment.” Living beings residing in the natural environment use their senses like sight, taste, touch, smell and hearing to perceive and comprehend different features of their surroundings. This sensory data is used by the brain at a subconscious level to create a model of the behavior pattern of things in the environment, which is then used in the interactions. This is exactly what a pet animal does as it gets introduced to a new owner and with time it gets comfortable around that person.
Imagine teaching perception to a mobile device where we consider the mobile device to be like a pet for its owner. The sensory organs in this case would be the sensors present on a device like accelerometer, proximity sensor, gyroscope, light sensor, etc. This sensor data is continuously collected by a device and can be used to build a dynamically evolving behavior profile for the primary user of the device.
Description:
Every user has a characteristic way of handling the mobile device including the angle at which she holds the phone, the height of the phone from the ground or the proximity of the phone from the body while she is using the device. Also, every user performs a certain set of activities in a day while carrying the mobile device with her. This could involve things like running/jogging, walking, sitting or lying down in a certain way or standing still. All these activities result into generation of massive amounts of sensor data over the period of a single day. This data can be used to teach the behavior pattern of the user to the device by employing a machine learning clustering model. The resulting clusters that get formed would be characteristic for the specific user and they would keep on evolving over time.
If the sensor data that gets generated at a given point of time lies in one of the clusters already formed, we can say that with a reasonably high probability, the device is able to recognize the user handling the device based on the behavior profile that it has built so far. A user recognition score can be generated which would shoot up in this scenario. On the other hand, if the current data doesn’t lie in one of the existing clusters, then there are two possible scenarios. Either someone other than the primary user is trying to use the mobile device or the primary user has performed an activity which has not been captured in the behavior model so far. In both the cases the user recognition score would go down. If it is the second case then the user shall be repeating similar activity in the future and hence the dynamic/online nature of the algorithm would ensure that the user profile would get molded to encapsulate this new behavior, i.e. the clusters would get updated accordingly. Due to this, the score won’t be going down if a similar data point shows up in the future.
Applications:
The series of user recognition scores generated in a day can be used to decide the degree of security or hardening required on a given mobile device. This information can be used by organisations to frame their policies regarding the usage of mobile devices containing critical corporate data, and by individuals who store valuable personal or bank details on their phones. It can also be used to assess the overall security of a device. Besides that, a continuous authentication mechanism can be created on a phone under which a functionality could be added that automatically locks the device if the recognition scores remain low for a sustained period of time.
Conclusion:
Thus, perception skills on a mobile device can be of huge benefit for an individual as well as an entire organisation. It is like personalization or customization of a device based on a person’s habits and usage patterns under which the device is in some sense dynamically interacting with the user and trying to learn and distinctly identify her, with security being the end goal. This feature can also be coupled with things like assessing different applications stored on the device and analyzing key configuration details regarding the operating system to provide a holistic view of the security posture of the concerned device.