Harnessing the Power of Biosignals
By Hugo Plácido da Silva and Ana Fred
NOTE: This is an overview of the entire article, which appeared in the March 2014 issue of Computer magazine.
Click here to read the entire article.
Biometrics refers to technologies that measure and analyze human body characteristics such as fingerprints, facial patterns, and iris recognition. Traditional biometric systems rely mostly on external anatomical traits that can be easily retrieved without user consent. It’s also possible to physically extract some of these traits from an authorized user’s body to gain illegitimate access to a system (see the article for some instances!). Aliveness detection ensures that a biometric signal is from a living source – ideally, its legitimate owner. There are, however, inherent limitations that have been difficult to overcome.
Biomedical systems should also be able to continuously recognize a user to ensure that he or she is the same person initially authorized to access the system. However, fingerprint readers can become occluded by dirt, lotions, and oil from fingers making it necessary to use repeated swipes to gain access. Facial and iris recognition systems require precise orientation to the sensor and are heavily influenced by ambient light and changes in pose or expression.
To address these issues, identity sciences are increasingly looking at biomedical signals such as cardiac activity, skin conductance, and temperature. These signals originate from a live source, are continuously available, and cannot be reproduced from latent impressions or remotely captured patterns. The article reviews some of these signals and devices to obtain them.
Behavioral biometric approaches leverage a range of actions or mannerisms that tend to be unique to a given person and are used sporadically. One of the most studied techniques is keystroke dynamics, such as the amount of time the subject holds down a particular key or moves between keys.
Biosignals are a promising solution to the limitations of current biometric systems, as they originate with a live source, enabling intrinsic aliveness detection, and are generally available in a continuous or near continuous manner. In particular, electrocardiography (ECG) is at the forefront of biosignal-based biometric approaches. Preliminary results indicate that ECG waveforms are measurably distinct across individuals and can serve as an effective modality for biometric verification even across several months.
The increasing availability of inexpensive, easy-to-deploy physiological sensors and hardware platforms should make it possible to seamlessly and cost-effectively integrate biosignal authentication with traditional fingerprint, face, and iris recognition, providing an additional security layer.
ABOUT THE AUTHORS
Hugo Plácido da Silva (hugo.silva@lx.it.pt) is a researcher at the IT – Instituto de Telecomunicações, and a PhD student at the IST – Instituto Superior Técnico, Portugal.
Ana Fred (afred@lx.it.pt) is a researcher at the IT – Instituto de Telecomunicações, and an associate professor at the IST – Instituto Superior Técnico, Portugal.
