Notre Dame researchers using iris recognition to improve accuracy of Somaliland election process

A biometrics expert at the University of Notre Dame is working with the Somaliland government to help the country use biometrics to improve the accuracy of its election process.

Since declaring itself an independent state in 1991, Somaliland has been working towards becoming a unique, multiparty democracy in the Horn of Africa.

University of Notre Dame biometrics professional Kevin Bowyer, along with his Ph.D. students Estefan Ortiz and Amanda Sgroi, have been working with the government to help realize this goal by developing a voter registration checklist free of fraud.

“One objective of the Somaliland government is to have truthful, respected elections,” said Bowyer. “Toward this end, they want to create a fraud-free voter registration list. They have turned to biometrics as a means to generate such a list.”

Somaliland performed a biometric voter registration exercise in 2008-2009, with fingerprints and facial recognition, and much effort was devoted to cleaning the voting register.

A report in 2010 by Electoral Reform International Services for the Somaliland Nationwide Electoral Commission (NEC) concluded that “this register is known to include a massive number of duplicates, perhaps around 30%, and the present biometric methods could not recognize these with the information offered.”

Based on the recommendation of its election experts, the Somaliland government contacted Bower’s research group to help it further explore the deployment of iris recognition as an alternative to fingerprint recognition.

The Bowyer group has published multiple articles on iris recognition technology, which played a huge role in convincing the NEC that iris recognition, when performed with the right equipment and procedures and with a focus on achieving high quality data, was a legitimate solution.

With the country’s voter registration legally required to be calculated by the end of 2014, Somaliland officials turned to Bowyer’s group to perform a test voter registration project using iris recognition that would be completed before the start of Ramadan (June 29).

Bowyer says that data acquisition was carried out over five days in two registration centers and transferred electronically to the research group at Notre Dame, where it carried out the iris recognition evaluation.

The researchers analyzed 1,062 trial voter registrations to determine the quantity of duplicate registrations. Each record contained two iris photos, for the left eye and the right eye.

Using automatic matching techniques, the Notre Dame team was able to rapidly browse through a set of 2,124 iris photographs to identify 450 duplicate registrations. They then manually inspected a small quantity of final results that have been ambiguous mostly based on the automatic matching, which successfully found another seven cases of duplicate registration.

The researchers submitted this record of 457 cases of duplicate registration to the Somaliland NEC, along with a technical report that detailed how they performed its analysis with recommendations on preserving and improving overall picture quality.

Elections expert Roy Dalle Vedove, who collaborated with the Somaliland NEC in improving the accuracy of the voting register, and NEC chairman Mohamed Ahmed Hirsi Gelleh were both very satisfied with the results.

Somaliland will continue to develop a new national, biometrically-validated voting register for use in future elections.

Article Topics

biometrics  |  elections  |  fingerprint biometrics  |  iris recognition  |  Somaliland  |  University of Notre Dame