ISO/IEC Joint Technical Committee Starts Deliberations on International Standardization of “Biometric performance estimation methodologies using statistical model” Proposed by Japan
- New methods for evaluating accuracy of biometric authentication making use of extreme value theory* -
September 23, 2020
The ISO/IEC Joint Technical Committee 1 (hereinafter referred to as the “JTC 1”) reached an agreement to commence deliberations on international standardization of “Biometric performance estimation methodologies using statistical model” based on a proposal submitted by Japan.
Successful establishment and publication of this international standard is expected to help industries to reduce the cost incurred in accuracy evaluation of biometric authentication devices. As people are requested to introduce new lifestyles and behaviors into their everyday lives, biometric authentication devices will be used more frequently. As an approach to address this situation, this proposed standard is expected to contribute to expanding ways of using biometric authentication and facilitating the popularization of such authentication at home and abroad.
1. Purpose and the background of the proposal
As the utilization of IT is becoming popular in a variety of business activities and consumers’ everyday lives, industries have been required to effectively provide enhanced safe, secure and meticulous services for users. To this end, industries have begun to make use of highly accurate biometric authentication, i.e., personal authentication by the face, fingerprint, vein, iris, voice, signature and gait.
ISO/IEC 19795 series (biometric performance testing and reporting) stipulate methods for evaluating the accuracy of biometric authentication. As the international standard organizations recommend industries to satisfy the standards, many industries have been applying them in manufacturing biometric authentication devices. Meanwhile, against the backdrop of advancing biometric authentication technologies and enhancing the accuracy thereof, industries need a significant number of samples when evaluating such technologies and they often bear the burden of excess costs for accuracy evaluation of the technologies, in particular, for the research and development of biometric authentication, performance testing of services, and assessing and evaluating systems and equipment.
In light of this situation, the Japan Automatic Identification Systems Association and its member companies with expertise in biometric authentication technologies jointly discussed and developed new methodologies for assessing the accuracy of biometric authentication with a smaller set of samples. Based on the outcomes of this effort, Japan National Body submitted a proposal for developing a new standard to the Working Group 5 (WG5: Biometric testing and reporting) of JTC 1/SC 37: Biometrics through a project which METI has commissioned to a third party since FY2019.** In response, the SC 37 international conference held in July 2020 unanimously approved the proposed work item to proceed to the working stage of the development of the new standard. Subsequently, JTC 1/SC 37/WG5 decided to start full-fledged deliberations on standardization from September 2020, aiming to publish the new international standard in FY2023.
*1. The term “extreme theory” refers to statistical methodologies for estimating the probability of the occurrence of rare events. For example, extreme theory is applied to assessing extreme climates or evaluating the potential risks of financial products.
**2. METI Project for Accelerating Strategic Development of International Standards: Activities for Development of International Standards in the Fields involving Governmental Strategies (international standardization for methods for evaluating the accuracy of biometric authentication which facilitate the evaluation of such accuracy in terms of security in cashless transactions), which METI is scheduled to conduct from FY2019 to FY2021.
2. Details of the proposal
In the process of biometric authentication, characteristics extracted from biometric features, e.g., face, fingerprints, veins and iris of the target person, are compared with registered templates (original data for comparison) in order to calculate the comparison scores (degree of similarity). In addition, setting a threshold value for determining whether the subject is the target person or not enables the calculation of a false acceptance rate,*** and this facilitates the evaluation of the performance of biometric authentication devices and systems.
Using the conventional evaluation methods based on ISO/IEC 19795 series, in which a general probability distribution is expected, industries need to collect an enormous number of samples since they need to measure false acceptance events which rarely occur in highly accurate biometric authentication devices for evaluation. In contrast, the proposed international standard submitted by Japan National Body this time is to specify alternative methodologies for evaluating biometric devices with a smaller number of samples by applying the extreme value theory.
According to the results confirmed in some model cases, the proposed methodologies successfully estimated the false acceptance rate for the database of 6,000 biometric samples by using only 1000 samples, i.e. one-sixth of the total samples. The estimation demonstrated a very good fit to the false acceptance rate calculated by using all the empirical data, which proved the effectiveness of the application of the extreme value theory in this field. JTC 1/SC 37 P-member countries are expected to start productive discussions to assess the feasibility of this statistical approaches from a technical point of view and include the proposed novel methodologies to the international standard.
***3. The term “false acceptance rate” refers to the proportion of biometric transactions with false biometric claims erroneously accepted
3. Expected effects
Successful publication of the international standard based on this proposal is expected to contribute to the reduction of the costs incurred in the accuracy evaluation of biometric authentication devices, the expansion of the biometric authentication and the facilitation of the popularization of biometric authentication at home and abroad, which will contribute to the realization of enhanced safety and security. Moreover, the proposed standard is also expected to enable industries to smoothly evaluate the performance of biometric authentication devices while assuming the use of such devices under new conditions, e.g., facial authentication of a person wearing a face mask. As people are requested to introduce new lifestyles and behaviors into their everyday lives to address the spread of the novel coronavirus disease, this proposed standard is expected to encourage industries to address increasing opportunities for using biometric authentication devices as part of working style reform (teleworking), educational reform (online classes) and lifestyle reform (medical care, care-receiver monitoring and unmanned stores).
Division in Charge
International Electrotechnology Standardization Division, Industrial Science and Technology Policy and Environment Bureau