IA Toolbox

This is a research tool for implicit authentication. It can capture all the sensors information including : lightmeter reading, motion, accelerometer, location (longigude, latitude), time, device id, device number, touch even (strength, time and etc.) and store them into SQLite cache database. You can specify the size of this database by input the number of lines of database from the app. The default value is 500. It can support at least 4000 lines but I recommend you to use no more than 3000 to keep the app running smoothly.

You can also export the cache database to .csv file. Such file can be used for feeding to machine learning tool box such as Mallet, Stanford topic model and Matlab. I will also upload the java file to further trim the file in order to suit the Mallet.

The other property of this toolbox is it can support dynamic sampling. Since to sample all sensors data to database must consume lot of energy. If you consider the battery usage, please check the check-box in the app. It will automatically calculate the JS-distance between training data and current "stride" data to achieve best sample rate.

Please also notice there is a yellow progress bar inside the app. It indicates the distance of current user and the historical user. The maximum value is 10, which indicates the current user behaviors abnormal comparing with the previous behavior history. At that time the sample rate will be high. In contrast, if the value is low, it means the device think you are the same user and sample rate will be low. For more detail, please refer my paper.

If ind any bugs, please leave me a message, I will fix it ASAP. Thank you