WatchAuth dataset (extended)

We collected inertial sensor data on a smartwatch as users performed activities. The data contains the tap gestures (n=31) as users performed tap gestures against payment terminals and activity data (n=9) as users performed out-of-lab activities (walking, commuting, in-store, or a combination of them). Each user folder contains 1 nfctimstamper.csv file, 1+ --watch-sensors.csv file(s), and 0+ --.csv files. These types of files are described as follows: - nfctimstamper.csv This file contains the in-lab timestamps collected for the purpose of segmenting the watch sensor data. The first column is the human-readable timestamp, the second is the UNIX timestamp, and the third contains the data read from the NFC tag that triggered the entry. The NFC tag is either 'userWTC' which was attached to the smartwatch or something containing 'TAP' which shows which terminal is being interacted with -- that is, the 'TAP' tag tells the system which terminal is being used and then each subsequent 'userWTC' is the timestamp of the NFC connection established during a tap gesture as the user interacts with that terminal. The third type of tag is of the form 'ATK--', which timestamps an impersonation attempt made by the user having watched a video clip of performing . - --watch-sensors.csv These files contain the in-lab inertial sensor data collected on the smartwatch. Use the nfctimestamper.csv file to segment this data into tap gestures. - --.csv These files contain the out-of-lab inertial sensor data collected on the smartwatch when the user was performing the . The following users participated in both the first and second runs of the experiment, roughly 18 months apart. These datasets can be used to investigate the time stability component of the tap gesture biometric. In the first run, they had the userIDs in the first column below; in the second, the second. 001 -> 020 008 -> 025 009 -> 027 010 -> 033 014 -> 037 018 -> 022