MINEX III FAQs

Q: How do I participate?

A: Follow the instructions on our participation page.

Q: How much does it cost to participate in MINEX III?

A: Participation is free.

Q: Do I need to be seeking PIV compliance to participate?

A: No, but you still must follow all instructions on the participation page.

Q: May I participate anonymously?

A: No, any software submitted will be listed along with the organization's name.

Q: May I participate but choose to have my results not posted publicly?

A: No.

Q: Do I need to be selling a commercial product in the US to participate?

A: No, any individual, company, or organization from most countries may participate. Check the BIS Entity List to see if you are not allowed to participate.

Q: I've already participated in MINEX, can I submit my new algorithm?

A: You may submit improved submissions once every 90 days. Each resubmission requires a new application. Each submission requesting PIV compliance requires a unique CBEFF Product ID.

Q: May I scan or fax my application instead of or in addition to mailing it?

A: NIST will start the MINEX III test if you send a scanned or faxed version of your application.  Your physical wet signature application will still need mailed to us.  

Q: When will my organization's name be posted to the MINEX website?

A: Your participation may be made public any time on or after receipt of a software library from your organization.

Q: May I submit only a template generator?

A: Yes.

Q: May I submit only a template matcher?

A: No. This is because in the typical operational access control scenario (the scenario tested in MINEX III), a template is generated for verification by the matcher supplier's template generator.

Q: May I submit a software library without a IBIA-assigned CBEFF Product ID?

A: No. You can register a new CBEFF Product ID with the IBIA (https://www.ibia.org/cbeff/). NIST hosts another fingerprint test, PFT-II, that does not require a CBEFF Product ID or many of the restrictions of NIST SP 800-76-2.

Q: May I participate if my validation package is not reproduced exactly at NIST?

A: No. Previously, the answer was "maybe." In NIST's experience, validation differences are a sign of software defects. In some cases, floating point math had caused inconsistencies in validation packages. However, almost all of these differences have been traced back to running on different patch levels of Windows or differently configured build-time settings. With MINEX III, NIST is mandating the use of a singular free operating system to prevent these inconsistencies and enable reproducible results.

Q: May I have a portion of the dataset for algorithm training?

A: No. The images contained in the MINEX III dataset are protected under the Privacy Act (5 U.S.C. 522a) and will be treated as Controlled Unclassified Information, Sensitive but Unclassified, or Law Enforcement Sensitive. NIST does provide numerous public datasets as well as a validation dataset, any of which may be used to test and train your algorithms prior to participating in MINEX III.

Procedure

Q: CentOS 7.2 is old, I want to use CentOS X.Y.

A: The operating system is standardized to ensure reproducible operation between interoperable submissions. Every so often, we will evaluate updated versions of the operating system to confirm that submissions operate reproducibly and upgrade the test operating system.

Q: Which products must my product interoperate with to be considered PIV compliant?

A: The PIV application is intolerant of non-interoperable pairs. Therefore, your submission must interoperate within the thresholds specified in NIST SP 800-76-2 with all existing submissions that have been shown to be compliant.

Q: Why does the MINEX III API require production of a template for all input images, regardless of content?

A: Ordinarily when an algorithm is given a poor input image, it may fail to produce a template. This is often referred to as a failure to enroll. In MINEX III, such cases must produce a template, referred to as a null template. The null template is a template that contains the INCITS 378 header and a single finger view with zero minutia. The MINEX III test requires a template in all cases in order to correctly account for such failures. A matcher that encounters a null template in a genuine comparison should likely produce a low similarity score, increasing the FNMR for genuine comparisons and decreasing the FMR for impostor comparisons.

Q: What is sum-rule fusion?

A: Sum-rule fusion is the process of summing the similarity scores produced by a template matcher when matching against two separate single-finger templates. The MINEX III API matching function only performs single finger matching. The sum-fusion process is used to estimate recognition error rates for a verification process in which two fingers are used in each attempt. This fusion guideline is outlined in NIST SP 800-76-2, Annex A, Section 4. MINEX III uses the mean of sum-rule fusion.

Q: INCITS 378 allows for encoding of minutiae quality, but MINEX does not. Why not?

A: This was true for Ongoing MINEX, but is no longer true in MINEX III. If you calculate minutiae quality, you may encode the values [1,100]. If you do not calculate quality, you may continue to encode 0. MINEX III template matchers are expected to successfully match templates regardless of minutiae quality.

Q: Why does the MINEX III not include three-way testing as conducted in the MINEX04?

A: The three-way scenario is not usually the most operationally relevant case. Typically when performing verification, a fingerprint is freshly captured, a template is generated by the matcher supplier's generator, and that template is then matched against a stored enrollment template. MINEX III targets this most common case — where participant's X template matcher matches templates created by generators from participant X and Y. The full three-way test is expensive to compute as the number of interoperable submissions increases. The scenario is not mandated by NIST SP 800-76-2.

Reporting

Q: Will NIST report results if my algorithm does not achieve PIV compliance?

A: Yes.

Q: Will NIST remove results if my algorithm performs poorly?

A: No, but you are encouraged to submit a new algorithm after 90 days to increase your accuracy.

Q: Why do the error rates reported in MINEX III differ from those from the same products used in previous MINEX tests?

A: MINEX III employs a larger sample of the POEBVA fingerprint population than used in both Ongoing MINEX and MINEX04.

Q: Will you send me statistics on how my algorithm performed?

A: Send us an e-mail if the information is not in your report card and we'll consider it. We're always looking for ways on how we can provide the most useful feedback on your algorithms so that you can move your products forward.

Q: What do I do if my algorithm has been decertified?

A: Resubmit an improved algorithm. We will work with you to get your algorithm certified once again.