Test results of the algorithms of Russian biometric companies in the world market

    In Russia, they are discussing the creation of a mega National Biometric Center with a database size of 100-150 million records. And the draft law on mandatory biometric registration has already been submitted to the State Duma . Since all of this, theoretically, is obliged to work on patriotic equipment, that is, for reasons of information protection, it is desirable to have brains and equipment geographically located inside the country, I think it will be interesting for you to get acquainted with what is good, theoretically, with us it will turn out.

    More recently, the two-year testing of pVTE-12 (Fingerprint Vendor Technology Evaluation 2012) at the US National Institute of Standards (NIST) has completed.



    The purpose of testing is to assess the real capabilities of fingerprint identification systems today. This is the largest testing (large databases and data types), which lasted for two years.

    Test data is divided into three groups:

    A - fingerprints of two index fingers without segmentation of 1.6 million records.
    B - ten fingerprints, in the group 4 + 4 + 2 (this is the removal of four fingers + thumb from each hand) with segmentation. The volume of the database is 3 million records.
    С - rolling and contact fingerprints of ten fingers in the 4 + 4 + 1 + 1 group with segmentation. The volume of the database is 5 million records.

    Task:for a limited amount of time, search for information in the "one to many" mode. Companies that did not meet the time limit were removed from the test. Therefore, all companies have prepared new high-speed identification algorithms.
    The protocol recorded not only the accuracy of the comparison but also the speed. Therefore, for testing companies provided two versions of the algorithm: fast and slower.

    Of the 22 two companies, only 18 were able to pass at least one test. These are almost all the biometric companies known in the world (NEC, 3M, Safran, etc.) as well as two Russian companies - Sonda, Papillon and several other new less well-known ones.

    Код компании	                Наименование			Тесты
    C				afis team			A, B, C
    D				3M Cogent			A, B, C
    E 	 			Neurotechnology 	        A, B, C
    F				Papillon			A, B, C
    G 		 		Dermalog			A, B, C
    H				Hisign Bio-Info Institute       A, B, C
    I				NEC 				A, B, C
    J				Sonda 				A, B, C
    K				Tiger IT 			A
    L				Innovatrics 			A, B, C
    M				SPEX				A, B, C
    O				ID Solutions 			A, B, C
    P				id3				A
    Q				Morpho 				A, B, C
    S				Decatur Industries 		A, B, C
    T				BIO-key 			A
    U				Aware 				A, B, C
    V				AA Technology 			A, B, C
    


    The most important test is the C test, based on 5 million records. By accuracy, the first three places were taken by giants - NEC (Japan), Morpho / Safran (France) 3M Cogent (USA).

    And our company Sonda (severe Chelyabinsk!) In the search mode for contact prints by rolling took 4th place!
    In search mode, contact by contact and rolling by rolling - 5th.

    The error probability is at the level of 0.3% and differs from the group of leaders by about half, however, in terms of speed of comparison and the length of the mathematical code, Sonda is superior to the leaders.

    As for Papillon, it’s very sorry. Apparently, the guys did something wrong with the algorithms and the probability of error from the group of leaders differs by 40-100 times and ranges from 5 to 20%, which in practice, of course, is unacceptable.



    Based on the test results, NIST identified five companies that provided optimal algorithms taking into account three factors: identification accuracy, speed of comparison and required resources: Sonda (Russia), Innovatrics (Slovakia), Morpho (France), IDSolutions (USA) and AA Technology (China) ) Following



    this event, I would like to receive a comment on the test results from our local biometrics specialist ZlodeiBaal .

    Full test report: nvlpubs.nist.gov/nistpubs/ir/2014/NIST.IR.8034.pdf

    Also popular now: