Data61 develops technology to use a person’s gait for powering wearable devices and for authentication
Researchers from CSIRO’s Data61 have developed a new technology which uses a person’s gait, that is the way they walk, to power wearable device and can be used as a new authentication method, which could replace passwords, pins or fingerprints.
Small sensors called accelerometers can currently be used to capture an individual’s gait in terms of motion and velocity. There has been a lot of talk about accelerometer-based gait recognition for mobile healthcare systems during the past few years.
It has several advantages over other biometrics when they are being integrated with wearables. Although fingerprint and face have been proposed for user authentication on smartphones, fingerprint readers and cameras are not always available on wearable devices such as pacemakers and smart watches. In contrast, walking is a daily activity. Moreover, using gait can be measured in a non-intrusive fashion.
The major bottleneck in development and adoption of such systems is that it requires continuous sampling of accelerometer, which reduces battery life of wearable sensors. The high power consumption of accelerometer sampling, typically on the order of a few milliwatts, also makes it challenging to adopt gait-based user authentication in wearables, which are often resource-constrained. Some devices like smartphones have large batteries but this poses a tough challenge for wearables like Implantable Medical Devices (IMDs) which are expected to be used long-term and where battery replacement requires surgical intervention.
In the paper, the researchers mention that a vision for wearable devices is to be battery-free or self-powered and kinetic energy harvesting (KEH) solutions for powering the devices is being investigated. However, the amount of power that can be practically harvested from human motions is insufficient to meet the power requirement of accelerometer for accurate activity recognition. In fact, it is 2-3 orders of magnitude smaller than what is required for the devices. Though power consumption of sensors has been largely reduced in recent years due to Ultra-Low-Power electronics, the researchers believe that in the near future energy harvesting will be used to augment or substitute batteries.
To overcome the power drainage problem, researchers from CSIRO’s Data61 combined gait recognition with KEH, which translates a person’s motion into electrical energy and improves battery life. They used an individual’s unique energy generation pattern as a form of authentication, instead of looking at an individual’s unique movements , to bridge the gap between the energy from KEH and that required to monitor the gait itself. The idea was that if humans have unique walking patterns, then the corresponding patterns of harvested power from KEH should be unique too.
Data61 researcher Sara Khalifa said, “By applying both techniques we have developed a way to achieve two goals at once - powering devices and the ability to verify a person’s identity using a wearable device by capturing the energy generated from the way they walk.”
The researchers conducted a trial on 20 users to evaluate the security of KEH gait authentication. Data was collected from each user using two different settings from various environments. Users walked in several environments including indoor on carpet and outdoor on grass and asphalt terrains to capture the natural gait changes over time and surfaces.
In the trials, KEH-Gait achieved an authentication accuracy of 95% and reduced energy consumption by 78%, compared to conventional accelerometer-based authentication techniques.
The KEH-Gait system was also tested against ‘attackers’ who attempted to imitate an individual’s motions. The analysis found only 13 out of 100 imposter trials were wrongfully accepted by the system as genuine trials.
However, the technology achieved approximately 6% lower accuracy compared to accelerometer based gait recognition. But then the researchers demonstrated that authentication accuracy can be increased to that of accelerometer-based gait detection by using a method, called Multi-Step Sparse Representation Classification (MSSRC), which exploits the information from multiple steps.
Group Leader of the Networks Research Group at Data61 Professor Dali Kafaar said highlighted the benefits of the KEH-Gait approach compared to passwords, pins, signatures and finger prints, saying, “It’s more secure than passwords because the way we walk is difficult to mimic. Since the KEH-gait keeps authenticating the user continuously, it collects a significant amount of information about our movements, making it difficult to imitate or hack unlike guessing passwords or pin codes.”
Data61’s privacy and authentication research team is also exploring other more secure and implicit continuous authentication techniques such as unique breathing patterns and distinctive behavioural biometrics from the way users innately interact with their devices.
Read the research paper here.