Authentication is a vital tool required to ensure security of your private data – well, unless you disconnect all your equipment from the internet and lock it in an underground bunker. Jokes aside, once you authenticate/unlock your device, it becomes potentially vulnerable whenever someone else takes this device into his (or her) own hands. How to avoid such danger?
One answer is very simple – not to lend your electronics to nobody even for a brief periods of time. Obviously, this is not very convenient as there could be many situations where you would have to leave your gadgets unattended for a very short duration. What if authentication could be done automatically, right after you pick your phone, tablet or any other digital portable piece of equipment into your hands?
It could sound similar to fingerprint-based biometric identification, but in fact it isn’t the same thing. You consciously have to place your finger on top of the fingerprint reader to make it work. Suppose you could make the entire surface of your smartphone capable of scanning your fingerprints, but that probably would not be a very smart or cheap solution from the technical point of view.
There could be another way: to use a continuous authentication based on user’s hand movement parameters. This idea was investigated in a study recently published on arXiv.org. In this study, a new biometric modality for continuous authentication of smartphone users has been proposed.
According to the presented concept, this technology uses accelerometer, gyroscope, and magnetometer readings to provide a continuous and undisturbed monitoring and capture of subtle user hand movements and orientation patterns which are generated when a user taps the screen or makes other common actions. A new biometric modality was correspondingly titled Hand Movement, Orientation and Grasp (HMOG) to denote main aspects used for personal identification.
The authors of the study argue that stability of the object held in user’s hand and repetitive nature of precision-demanding actions like tapping the screen are sufficiently distinctive features which retain their unique character despite a wide variety of ways to hold phones or tablets and therefore could be used to provide an additional level of interaction with portable electronic devices.
“Studies in ergonomics, biokinetics, and human-computer interaction have reported that handgrip strength strongly correlates with an individual’s physiological and somatic traits like hand length, handedness, age, gender, height, body mass, and musculature”, note the researchers. Motivated by these facts, they devised 96 individual HMOG features and evaluated them according to performance parameters required to accomplish the continuous user authentication and biometric key generation during a process of touchscreen-based typing.
Two separate cases related to typing performance – under walking and sitting conditions – were evaluated. A dataset of HMOG features has been collected from 100 users who typed on the smartphone. They had to answer three questions, typing at least 250 characters for each answer. Respective accelerometer, gyroscope and magnetometer output signals were evaluated according to their user authentication performance. The scientists note that HMOG features extracted from accelerometer and gyroscope signals outperformed those extracted using magnetometer readings.
Also, when HMOG features have been augmented with tap characteristics, such as tap duration and contact area, it was possible to lower equal error rates (calculated over the so-called authentication latency time window, which was 20 seconds during experiments). For example, for 20-second authentication latency, combined HMOG, tap and keystroke dynamics allowed to achieve the equal error rate of 6.92% for walking and 10.20% for sitting positions.
“Our results demonstrate that HMOG is well suited for continuous authentication of smartphone users, because the performance of HMOG improves during walking – a common smartphone usage scenario”, the team wrote. They also note that the energy consumption increase associated with sample collection and feature extraction was relatively small (less than 8% energy overhead when sensors were sampled at 16Hz), and therefore a newly proposed authentication method could be well-suited for devices with constrained energy resources, which include most of portable electronics.
Written by Alius Noreika