Overview#Presentation Attack Detection (PAD) is a framework through which Presentation Attack events can be specified and detected so that they can be categorized, detailed and communicated for subsequent decision making and performance assessment activities.
One of the challenges in implementing biometric authentication systems is deploying safeguards to mitigate against Presentation Attacks. This is especially true in uncontrolled environments where there is not an operator monitoring the placement or capture of a biometric. There are multiple ways to mitigate the risk of a successful presentation attack. The class of methods created to directly counter these attempts at the biometric sensor is known as Presentation Attack Detection.
Presentation Attack Detection is defined as "automated determination of a Presentation Attack".
Presentation Attack Detection categories include
Error rates for Presentation Attack Detection are expressed as classification errors. A Presentation Attack Detection subsystem may incorrectly classify a bona fide Biometric Sample as an attack Biometric Presentation or vice versa. To measure the strength of function for Biometric Presentation, the error of interest occurs when a Presentation Attack Detection subsystem fails to detect an attack Biometric Presentation, allowing an attacker positive access to an authorized user’s account.
At the Presentation Attack Detection subsystem level, this error rate is known as the Attack Presentation Classification Error Rate (APCER) and is defined as the proportion of attack Biometric Presentation using the same Presentation Attack Instrument (PAI) Presentation Attack Instrument Species (PAIS) incorrectly classified as bona fide Biometric Presentation at the Presentation Attack Detection subsystem in a specific scenario.
At the biometric system level, this error rate is called the Imposter Attack Presentation Match Rate (IAPMR), which expresses the ratio of imposter attack Biometric Presentation, all made with the same material, that are successfully Authenticated (at a fixed operating threshold for the matching algorithm). In other words, Imposter Attack Presentation Match Rate is the measure of the ability of a spoofed biometric Sample to not only bypass the Presentation Attack Detection subsystem but to be authenticated as a bona fide user.