A biometric recognition and ultimately a successful authentication depend upon the robustness of the selected machine learning algorithm used within the solution but also on (as we discussed in the Feature extraction section of this chapter) the sample size.

In addition to these requirements, it is important to consider the quality of the samples as well as the type. Poor image quality, for example, can significantly impact the accuracy of a biometric authentication. We also briefly mentioned using behavioral traits as part of a biometric signature.

Generally speaking though, physiological characteristics (such as a fingerprint or facial picture, for instance) are always the most static, showing little dissimilarity over time, while behavioral characteristics (that is, a gait or cadence) can and usually do experience variations, and can be prejudiced by external factors or by particular emotional conditions such as stress or strong psychological impacts.

Interesting behavioral characteristics already in use by some biometric authentication solutions include vocal imprints, writing and/or typing style, movements of the body, the style and the trend of walk, and so on.