NZ government asks for help understanding facial recognition

The government of New Zealand has issued a request for information, as it aims to better understand facial recognition amid concerns about bias, particularly in use by law enforcement.

Like many places aiming to harness the potential of emergent biometric technology, New Zealand is host to a debate over how to regulate facial recognition, and how to address fears that it will lead to discrimination, especially against the Māori and Pasifika populations. With this call for info from the Department of Internal Affairs (DIA), the government is not requesting quotes or proposals, but is “seeking to understand ‘the what we need’ and matters such as: what is available in the market, which suppliers are active in this area and how suppliers address the various challenges pin this domain.”

In short, this is the get-to-know-you phase of a relationship that aims to establish a long-term commitment to biometric systems that provide secure identity verification and fraud prevention.

A report from RNZ looks at research which shows training facial recognition algorithms on synthetic identities may provide the beginning of an end to worries that the technology is inherently racist.

Researchers at New York University trained facial recognition algorithms on 13 million synthetic images of diverse faces, in order to reduce bias and improve accuracy.

These kinds of efforts will only improve what is, statistically, already a highly accurate process.

Some facial recognition systems are down to 0.02 percent inaccuracy in benchmark testing by the U.S. National Institute of Standards and Technology (NIST).

Even a Māori reference group has stated that the current accuracy of facial recognition algorithms is “pretty good” and likely to improve.

Concerns about bias are likely to continue, as debates about structural racism and digital inclusion remain hot. But testing will help further build the case for facial recognition. The DIA has enlisted BixeLab to test its Identity Check system’s biometrics and liveness detection for bias. Bixe confirmed this in an email to Biometric Update.

Article Topics

accuracy  |  biometric bias  |  biometric testing  |  biometrics  |  BixeLab  |  facial recognition  |  New Zealand