Company develops biometrics for use by Brazilian electoral system07 de agosto de 2018
By Suzel Tunes | FAPESP Research for Innovation – In Brazil’s next general elections scheduled for October 2018, more than 2,800 municipalities across the country will use biometrics to identify voters. According to data from the Electoral High Court (TSE), more than 87 million people had registered digital fingerprints by July.
The TSE aims to complete fingerprint registration of the entire electorate – more than 147 million people – by 2022. In this venture, it is relying on biometric software developed by Griaule.
Griaule is the only supplier of biometric individualization to the TSE. Its contract was awarded by competitive bidding. It was founded 16 years ago and has established itself with support from the FAPESP Innovative Research in Small Business Program (PIPE). It has also won federal funding from the Brazilian Innovation Agency (FINEP) and the National Council for Scientific and Technological Development (CNPq).
João Pedro Weber, Griaule’s project manager, says the business was started in 2002 when founder Iron Calil Daher, an electrical engineer, moved from Goiânia to São Paulo State to incubate Griaule at the University of Campinas (UNICAMP). “Proximity to the university enabled us to bring in some highly talented people to join the team,” Weber recalls.
Griaule incubated between 2002 and 2005, a period that was “crucial for us to build a solid base” from both the technical and the business standpoint, he explains. The first project undertaken with support from PIPE was submitted directly for Phase 2, which requires the development of an innovation. It began in 2004 with the aim of upgrading the automatic fingerprint identification system (AFIS) already marketed by the firm.
Recognition came swiftly. In 2005, the firm won the FINEP Award for Technological Innovation in the category Small Enterprises in the Southeast Region. A year later, it achieved international visibility by winning the Fingerprint Verification Competition with the AFIS that had the lowest error rate, standing out among 150 competitors.
Griaule’s market position was strengthened further by the successful completion of two other PIPE Phase 2 projects: “Digital human face detection and recognition”, completed in 2009, and “Voice recognition: speaker verification”, completed in 2011.
The firm began with five people and now has a workforce of some 40, from trainees to postdoctoral researchers, 60% of whom are directly engaged in R&D in different areas. In 2007, it set up a unit in the United States, first in California and more recently in Morgantown, West Virginia. “The unit has workstations, conference rooms, and space for servers running our solution on a large scale. It contributes to sales and local support for the US market, using the model successfully implemented in Campinas,” Weber says.
According to Weber, the choice of an open architecture similar to that used by companies such as Google and Amazon has also contributed to the success of the firm’s products. “In several institutions, both public and private, a biometrics solution offered on a proprietary platform has made customers entirely dependent on the supplier,” he says. Griaule’s system gives the customer more autonomy and its workings are more transparent.
Griaule has more than 4,000 customers in 70 countries. In Brazil, it has a presence in various markets, such as the banking segment, where Diebold Nixdorf has integrated Griaule’s technology into the systems used by the leading financial institutions; the digital certification and retail segment, where its customers include Serasa Experian; and the civil identification segment, where they include the TSE.
The software acquired by the TSE from Griaule – which resulted from more than a decade of research with the support of government funding agencies and INOVA UNICAMP, the university’s innovation arm – is designed to store, validate and authenticate voters’ biometric data. “It guarantees the uniqueness of each citizen’s record with the aim of eliminating fraud in the issuance of voter ID cards and in voting itself,” Weber says.
Other customers are Caixa Econômica Federal (CEF), a state-owned bank that uses the technology in its branches and to control the disbursement of family allowances under the Bolsa Família social program, and Fundação Casa, a juvenile correctional institution run by São Paulo State, which uses the system to identify undocumented inmates, biometrically recording all ten fingerprints as well as photographs, tattoos, scars and birthmarks.
Although the AFIS platform is Griaule’s flagship product, it also develops systems for face, iris, voice and palm print recognition (the last including newborns), as well as latent search. Davi Gouvea Martins, another project manager at the firm, explains that a latent biometric is a fragment of a fingerprint or other variable used in crime scene investigation and police work. “Fingerprint or palm print fragments are often used in investigations, for example. The software checks these fragments against the database to identify a suspect,” Martins says.
According to Griaule’s specialists, each form of biometric identification has pros and cons that depend on the context in which it is used. “Iris and palm identification have high accuracy ratings, but the biometric collection devices deployed in these cases are often invasive and costly,” Weber says. “Bearing this in mind, fingerprint recognition offers the best cost-benefit ratio when implementing a biometric system. It’s widely accepted, and hardware vendors are steadily improving their solutions and lowering the cost.”
According to Martins, Griaule specializes in biometric identification software capable of running on different types of hardware, not just for collection but also for data processing, such as fingerprint readers, cameras, servers and databases already available in the marketplace. Given this focus on software, it is dedicated to the development of increasingly robust algorithms and secure systems capable of preventing fraud in the collection of biometric data via the use of false fingers made of silicone or latex. “The software identifies characteristics that can’t be reproduced by false fingers, such as blood flow and the light refraction that occurs when the skin is lit up,” Martins says.
Address: Avenida Romeu Tortima, 1448, Cidade Universitaria, 13083-897, Campinas (SP), Brazil
Tel: +55 19 3289 2108