Skip to content
A data set of children's and adults' interactions on mobile devices
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
Kids
Parents
README.md

README.md

KidsOnThePhone_dataset

A data set contains touch and sensors data of children and adults interactions on mobile devices.

This is the data set collected and used in the Kid on The Phone paper

Toan Nguyen, Aditi Roy and Nasir Memon, "Kid on The Phone! Toward Automatic Detection of Children on Mobile Devices", Computers & Security, Volume 84, July 2019, Pages 334–348.

Overview

This data set was collected in a study with a goal to develop algorithms for automatic detection of children on mobile devices. In our work, we devised several methods to distinguish children from adults based on behavioral differences while operating a touch-enabled modern computing device. Behavioral differences are extracted from data recorded by the touchscreen and built-in sensors. Our results showed that it is possible to achieve 99% accuracy and less than 0.5% error rate after 8 consecutive touch gestures using only touch information or 5s of sensor reading. If information is used from multiple sensors, then only after 3 gestures, similar performance could be achieved.

The data set had been created from 50 children and adults who interacted with off-the-shelf applications on smartphones. The data set includes touch and sensor data of 25 kids (age range 3.5--12, average 6.6) and 25 adults (age range 24--66, average 36), mostly parents of kid subjects. The data set in this repo is stored in two directories: Kids contains data of children subjects and Parents contains data of adult subjects. The data of each subject is stored in a subdirectory within these two directories has following structure.

-- SubjectID dir, i.e., Kid1

------Touch: subdirectory contains touch data of the subject. Under this directory, there are subdirs, i.e, 1499123735344_kid, where each subdir contains the data of a session. Most subjects only have one session.

------Sensors: subdirectory contains sensors data of the subject. Under this directory, there are subdirs, i.e, 1499123735344_kid, where each subdir contains the data of a session. Most subjects only have one session.

------listApps.json: This is a file that maps all the apps on the device to app IDs. The app IDs were used in sensor data (the last column).

Touch data format

In each session directory, the touch data is stored in a JSON file with format session_timestamp.json, i.e., 1499123737171.json. The touch data includes following fields.

    + dimX, dimY: the screen size of the phone used in the data collection. In our study, we used a __Nexus 6__ phone to collect data from all subjects.

    + gestures: a list of touch gestures captured in the session. Each touch gesture has following data:

      ```
          {
	      "12": "Swipe",  --> Gesture ID and gesture type
	      
	      "app": "com.google.android.apps.photos",  --> The name of the app on which the gesture was performed and captured
	      
	      "points": [
	        {"o":true,"Ts":[{"p":56,"t":1499123762947,"s":6,"x":1146,"y":1478},{"p":56,"t":1499123762969,"s":6,"x":1117,"y":1481},{"p":56,"t":1499123762977,"s":6,"x":1099,"y":1482},{"p":56,"t":1499123762986,"s":7,"x":1074,"y":1483},{"p":56,"t":1499123762993,"s":6,"x":1038,"y":1483},{"p":56,"t":1499123763001,"s":7,"x":988,"y":1485},{"p":56,"t":1499123763014,"s":6,"x":937,"y":1487},{"p":56,"t":1499123763020,"s":6,"x":886,"y":1494},{"p":56,"t":1499123763027,"s":5,"x":835,"y":1503},{"p":56,"t":1499123763036,"s":4,"x":784,"y":1519},{"p":32,"t":1499123763045,"s":2,"x":735,"y":1536}]}
	      	] --> A list of touch points
	       },
	```

where

	    + `o`: orientation of the phone. `true`: vertical, `false`: horizontal
    
	    + `Ts`: touch sequence. Contains a list of touch points where each touch point has:
    
	      + `p`: pressure
      
	      + `s`: size of touch finger
      
	      + `x`: x-coordinate, the x location of a touch point on the screen
      
	      + `y`: y-coordinate, the y location of the touch point on the screen
      
	      + `t`: timestamp in UTC

Note: To normalize pressure and size value to [0, 1], use pressure normalization scale provided by the device used in the data collection process. See details: https://source.android.com/devices/input/touch-devices#pressure-field. For our device, this pressure_scale is 0.013.

Please note that gesture ID may not start from 1 because initial gestures, which belonged to the experimenter who helped carry out a data collection session, were removed from the data of a subject.

Sensor data format

In each session directory, there are subdirs of sensor types. Each subdir contains data of a sensor type during the data collection session, including, acelerometer, gyroscope, gravity, linear acceleration, magnetic, orientation, and rotation.

In each sensor subdir, there are *.txt files that contain captured data of the sensor. Each filename is a timestamp when the file was created. Each file has 5 columns:

      + `timestamp` in UTC: this timestamp should be used when combining with touch data
  
      + `sensor_timestamp`: sensor internal timestamp
  
      + The next three columns are sensor measurements along three axes `x,y,z`
  
      + The last column is the `app ID` (which app the subject was using when the data was captured). The app ID can be mapped to an app name using the file `listApps.json` described above.

Kid subject age details

Subject ID Age
Kid1 9
Kid2 11
Kid3 9
Kid4 7
Kid5 9
Kid6 5
Kid7 7
Kid8 7
Kid9 4
Kid10 4
Kid11 7
Kid12 5
Kid13 6
Kid14 7
Kid15 8
Kid16 7
Kid17 7
Kid18 10
Kid19 12
Kid20 3.5
Kid21 5
Kid22 3.5
Kid23 5
Kid24 4
Kid25 4

Reference

Kindly cite following article in any of your publication that benefits from the this data set:

Toan Nguyen, Aditi Roy and Nasir Memon, "Kid on The Phone! Toward Automatic Detection of Children on Mobile Devices", Computers & Security, Volume 84, July 2019, Pages 334–348.

bibtex:

@article{nguyen2019kid,
  title={Kid on the phone! Toward automatic detection of children on mobile devices},
  author={Nguyen, Toan and Roy, Aditi and Memon, Nasir},
  journal={Computers \& Security},
  volume={84},
  pages={334--348},
  year={2019},
  publisher={Elsevier}
}

PDF: Kid on the phone paper

You can’t perform that action at this time.