How Polymatica helped to improve the quality of service for the Unified Biometric System
The Unified Biometric System (UBS) is a Russian federal digital platform that allows a person to be identified by their biometric characteristics. It is used both for identification at a distance, which is very important in the current situation, and for face-to-face recognition. The system is designed to improve both the security and convenience of operation.
After initial launch of the UBS, problems were discovered in the system operation: the duration of the process of registration of biometric data for people in bank branches was as high as 40-50 minutes, which is unacceptable for the service.
To resolve the problems strategic goals were set: to reduce the elapsed time of data registration at the bank to just 7 minutes, and to increase the rate of success of registration to 97%. It was decided to create a monitoring system to manage the problem so helping to systematize the existing data and make it possible to take decisions on future development of the UBS based on data which could be relied on.
It was necessary to receive and process data from three federal systems each with a different SLA and different data processing methods, as well as interfacing with the banking systems. At the same time, these data systems must be located in secure circuits with a high level of security for storage and transmission.
The system involved rapid processing of up to 1,500,000 transactions per day related to connection and use and presenting a visualization of the resulting aggregates in a set of dashboards with the possibility to drill down and obtain detailed information for specific cases as needed.
The system operator should have an option to create his or her own dashboard quickly and without any knowledge of programming. Such a dashboard could be added both for all the banks in the system and for any specific bank with its personal account.
The monitoring system of the UBS allows users to evaluate the effectiveness of the system operation and to find and eliminate the causes of any errors. Data collected by the monitoring system makes it possible to answer a number of strategically important questions. Examples include how new users are registered, how many people have already had biometrics stored, and what weaknesses there are in the current system. This information can be used to make strategic decisions on any changes necessary based on the data provided from the system.
The following software was used. To ensure the collection, processing and storage of initial data and the calculation of indicators provided;
○ Yarn — as a program execution tool;
○ HDFS — as a file system for storing data.
● Hive — as a tool for implementing SQL queries;
● Spark — as a streaming data processing framework;
● Logstash — as a prefilter and router for data transmission;
● Filebeat — as a data collector agent.
For the analytical platform:
● Polymatica Analytics — an engine to process Big Data efficiently;
● Polymatica Dashboards — a tool to design dashboards;
● Pentaho — a tool for ETL;
● Docker container to roll out on-Premise.
THE RESULTS AFTER WORKING WITH POLYMATICA
The system is already being used for banks, and there are plans to extend use to any services that require identification of a person’s identity, including within government services. For such mission-critical applications, it is necessary to ensure the high quality and stability of service delivery. The objective was to ensure the highest quality of service and use of a software stack of the best products on the market, designed to handle data of this kind, made it possible to improve the service quickly and using a relatively low level of resources.
Quantification of the project results is given below:
Kirill Pruntov, Data Manager at Rostelecom:
“The monitoring system helped us to see a picture of what was happening in real time and so the processes were promptly changed and corrected when needed.
After the introduction of this monitoring and analysis of errors, work was carried out with banks and the process of collecting biometric samples was adjusted, making it possible to reduce the number of errors in the third quarter compared to the second quarter of 2020: for photographic data – by 24%, for audio data – by 34%. Currently, the number of registration errors has decreased by 87%, and the registration time has decreased by 8%. The targets set before implementation of the monitoring system has been overfulfilled: the registration time has been reduced to 6 minutes (target was 7 minutes), the success of registrations has been increased to 98-99% (target was 97%).”
Improvement of the quality of the registration process requires coordination of the work of all participants in the process. The unique feature of this monitoring system provides both aggregate and detailed data for internal users at Rostelecom, as well as individual data provided to the bank’s personal account.
Internal users receive data more or less in real time and so can quickly respond to non-standard situations, for example, the occurrence of several dozen errors in the delivery of voice biometrics. External users – banks – have access to the same set of dashboard templates through the bank’s personal account, but for each, only the relevant data from their validated requirement is loaded according to the security policy.
Any user can create a new dashboard in a few minutes using the convenient designer package provided.