Case Study

MegaMatcher Automated Biometric Identification System (ABIS) and MegaMatcher Accelerator Extreme provided fast and accurate deduplication of 46.7 million multibiometric voter records, identifying more than 6 million duplicates and 900,000 under-age records. Working with the Independent National Electoral Commission of the Democratic Republic of the Congo, Neurotechnology completed the project in under two months.

The Democratic Republic of the Congo (DRC), in preparation for their 2018 election, wished to have as many voters registered as possible and to have those registrations be an accurate record of all potential voters. With the multibiometric data of 46.7 million voters in their collected database, the Independent National Electoral Commission (Commission Électorale Nationale Indépendante or CÉNI) worked with Neurotechnology to deduplicate the registration rolls. Using MegaMatcher ABIS, based on the MegaMatcher Accelerator Extreme multibiometric matching engine, more than 6 million duplicates and more than 900,000 under-age records were identified, and results were achieved in less than two months.

For this very large-scale project, CÉNI evaluated a number of solutions and chose Neurotechnology as both the technology provider and the integrator to help them attain a set of accurate voter records. Through various field operations, 46.7 million multibiometric enrollments were collected and Neurotechnology was tasked with verifying and deduplicating the resulting dataset.

Using the MegaMatcher Accelerator Extreme (MMA Extreme) matching engine, which can match up to 1.2 billion fingerprints and faces per second, Neurotechnology set up a MegaMatcher Automated Biometric Identification System (ABIS) turnkey system to perform the full all-with-all comparison of 10 fingerprints and facial biometric for each record, a matching magnitude of more than 46.7 million x 46.7 million.

• The customer: Astride the equator, in the center of Africa, the Democratic Republic of the Congo (DRC), is, by area, the largest country in Sub-Saharan Africa and the 2nd largest country on that continent. With a population of over 86 million, the DRC is the fourth most-populated nation in Africa and the 16th most populated country in the world.
• The need: Understanding that to have fair and just elections it was essential to verify the registered-voter base of 46.7 million, CÉNI worked with Neurotechnology to verify and deduplicate the dataset of multibiometric face and fingerprint records for each individual, ultimately finding more than 6 million duplicate and 900,000 under-age voter records.
• The challenge: With 2018 elections fast approaching, the DRC needed to deduplicate their voter database within a two-month timespan to ensure that all constituencies in the country would be fairly represented in parliament.
• The integrator: Neurotechnology implemented this project using proprietary technology to complete the system installation and deduplication process support on site.
• The solution: MegaMatcher ABIS is Neurotechnology's turnkey solution based on MegaMatcher Accelerator – providing high-speed, high-volume fingerprint, iris and face biometric identification and database deduplication for national-scale projects.

"In the voter registry cleaning process our goal is to find as many duplicates and other noncompliant cases as possible. Neurotechnology helped us achieve our goals and exceeded our expectations by starting and completing the deduplication process in record time, and they were able to identify millions of duplicates." Mr. Corneille Nangaa Yobeluo,
President of Commission Électorale Nationale Indépendante or CÉNI.

The DRC National Assembly is based on the number of constituency voters, so it was very important to be as accurate as possible when establishing the voter rolls. To ensure the accuracy of the voter database, the Neurotechnology team set up the MegaMatcher ABIS to process the multibiometric face + 10 fingerprints records of all 46.7 million registered voters using Neurotechology's MegaMatcher Accelerator Extreme matching engine in an all-with-all, simultaneous comparison – matching every record with every other record in the database.

The ABIS and deduplication process in a large dataset, such as that collected by the DRC, requires a robust system. While some biometric systems attempt to simplify processing and enhance speed by splitting deduplication by regions or by dividing people into groups (e.g. by age or gender), this reduction in complexity can also reduce accuracy. These reduced or sliced datasets can sometimes show impressive results for a number of different vendor solutions. But when the datasets are large, multi-faceted and addressed as a whole, it requires a more powerful system like Neurotechnology's MegaMatcher ABIS and MegaMatcher Accelerator Extreme.

As a number of elections have been delayed in the DRC, it was deemed important that the current process should move ahead quickly and, most importantly, accurately. In this case, with Neurotechnology as the integrator, the deduplication process – from setup, requirements checking, configuration and piloting through the final deduplication of the entire dataset – took less than two months.

Working with the total of 46,815,156 multibiometric voter records, Neurotechnology's software and operator team found 6,002,346 duplicate or multi-duplicate records, more than 12%+ of the entire set. The system found 300+ situations where individuals had registered for elections more than six times. In addition, this process identified that 902,290 registered voters were under-age, and thus ineligible to vote. These numbers are significant enough to impact the elections, considering that DRC constituencies are represented in Parliament based on their representative numbers.

The accuracy of this process was not based solely on the automated deduplication system, which did the all-with-all matching of fingerprint and face biometrics and age identification; it also included a cadre of nearly 300 operators that manually reviewed and confirmed the duplicate and under-age records identified by the system.

While the MMA Extreme setup was able to deal with the 46.7 million x 46.7 million matching challenge, due to its efficiency it still only used 11 servers and the entire MegaMatcher ABIS system easily fit within two racks, greatly simplifying maintenance while keeping associated costs at a minimum.

The DRC deduplication process took advantage of the modular architecture of the MegaMatcher ABIS (MM ABIS). The structure of this Neurotechnology system makes it configurable to different functional and performance requirements. The architecture can be logically separated into various services and applications.

The Management Service of the MM ABIS is designed to maintain system integrity, ensuring that individual operations are executed consistently. It is the core component of MM ABIS, responsible for orchestrating the work of all system services and managing the workflow of operations.

The Image Processing Service is based on MegaMatcher biometric template extraction technology. This component is used for biometric data quality assessment and biometric data extraction from fingerprint, face and iris images.

The Matching Service is a high-performance biometric engine which performs the following operations:

• Identity verification – matching a biometric template against a selected database record (1-to-1).
• Person identification – matching a biometric template against the whole database (1-to-many).
• Records deduplication – matching all biometric templates in the database against each other (many-to-many). The questionable matching results are resolved via the adjudication procedure, either automatically or by human operators.

"To create, deploy and use Automated Biometric Identification Systems at a small scale is relatively easy; complexity and major challenges appear with scale. With successful implementation of this project, our MegaMatcher ABIS and MegaMatcher Accelerator Extreme products proved to be suitable for large national scale deployments." Mr. Denis Kačan,
MegaMatcher ABIS product manager for Neurotechnology

The service is based on MegaMatcher Accelerator technology and is controlled by the Management Service. The Matching Service is a scalable unit. Multiple units can be arranged into a cluster to meet a specific project's performance and capacity requirements

The MM ABIS Database Service provides reliable storage for transaction related data. It can be part of a reliable cluster of multiple database instances. MM ABIS uses two databases – Main and Template.

The Main Database is accessed only by Management Service and is used for MM ABIS specific data:

• Transaction information (graphical and standard-based biometric samples, biographic data, contextual data, status, result, registration timestamp, execution timestamp, completion timestamp, etc.).
• User information (user names, roles, actions taken).

The Template Database, used by the Matching Service, is the primary storage area for biometric data. The records are stored in a Neurotechnology proprietary format to ensure high performance.

The Adjudication Module is intended to provide a comprehensive toolset for biometric experts to manually investigate biometric operations and make decisions about unresolved records.

The System Administration Module is designed to manage and monitor the MegaMatcher ABIS system using role-based access control and consists of the following parts:

• System Management – provides ability to manage the MegaMatcher ABIS system's state.
• System Monitoring – dedicated for visual MegaMatcher ABIS real-time execution monitoring.
• User Management – dedicated to create and revoke users as well as provide or restrict access permissions to particular MegaMatcher ABIS modules.
• System Configuration – provides the ability to define search engine criteria (such as matching thresholds for identification and verification).
• Report Generation – statistical reports, with filtering options, are provided for complete MMABIS action inspection.