Biometrics databases are often flawed – people will suffer for their inexperience

November 24, 2022November 25, 2022 Mizan Rahman 0 Comments

755 views

Spread the love

Biometric systems can have detrimental long-term consequences if they are not designed properly. Let’s take a look at a recent incoming request.

Incoming lead details

State Health Insurance Agency described their project as follows: our role is to provide technical support and infrastructure to meet the agency’s needs. Therefore, we need to develop a duplicate resolver to isolate already registered lives (Enrollees). These enrollees are from different local government areas of the state, including both urban and rural settlements with little or no network. Our Biometric Enrolment Device can work online and offline. We capture the enrollee’s primary, facial, and biometric data. Initially, we started with capturing ten fingers per enrollee, but the engineers later changed this to four fingers – two left and two right fingers. We did not buy any biometric search system, and a de-duplication system is urgently needed. The current enrolment database size is 50,000 using fingerprint biometric data in ANSI 378 template format. Our goal is to have 2M+ as we cover the entire state.

Response from our solution engineer

Welcome to biometrics. Without knowing anything, they are capturing ANSI templates!!! No one in the industry can perform fingerprint deduplication on ISO/ANSI templates except us. I would also like to know which fingerprints they capture from each hand. Given their poor understanding, I bet they are capturing thumbs and indexes – this will be very bad.

They can use our hyper-deduplication to clean up the database and our CloudABIS to perform one-to-many biometric matching on ANSI/ISO templates. We are the only company in the world that developed such capabilities – please don’t ask why we developed this!

Here is more information on M2SYS DeDuplication System

In a traditional deduplication project, many servers perform 1-N on all database records to isolate the duplicates. That’s why companies put in dozens of servers, and it takes a long time to duplicate a big database. Here is the math. If you have 1,000,000 fingerprints to DeDuplicate, to perform one-to-many ( 1 to N) on all records, the minimum required search operations = N^2/2 = (1,000,000 * 1,000,000)/2 = holy cow! This math is only for 1 million records. Now you know what all those servers are doing!

M2SYS invented a new algorithm for DeDuplication – we do not perform one-to-many operations to find the duplicates. Therefore, we use 87% less hardware and reduce the time by 95%. That’s not the only bragging right; we can work on the templates from anyone – IDEMIA, NEC, Neurotech, Innovatrics, and more. Customers can keep their existing AFIS/ABIS and use M2SYS DeDuplication on the side to clean up the database.