NSW Police has deactivated Cognitec’s facial recognition capabilities due to accuracy concerns, particularly its high false match rates for certain demographic groups. The police stated they will utilize other technologies for identification and investigation purposes. Ongoing discussions emphasize the importance of responsible use and guidelines for emerging video analytics technologies.
The New South Wales (NSW) Police have deactivated the facial recognition capability within their PhotoTrac Suspect Identification System, which utilized technology from Cognitec, a German company. This decision, made in February, was confirmed in response to an inquiry by the NSW Parliament regarding budget estimates. The system compared CCTV images of potential suspects with mugshots to assist in identification.
Concerns about the accuracy of Cognitec’s facial recognition software have prompted the deactivation. Media reports highlighted that the technology misidentified West Africans significantly more often than Europeans, with some instances showing a false match rate as high as 0.538% for West African females aged 65 to 99, as noted in the Face Recognition Technology Evaluation report.
The NSW Police Force employs various facial recognition systems. Despite the existence of updates from Cognitec, the police did not acquire them due to the historically high error rates associated with legacy algorithms across diverse demographic groups. Notably, the police also adopted an open-source facial recognition system named FaceNet, internally branded as NYX, in 2018.
Facial recognition technology has been instrumental for NSW Police, generating leads more than 560 times in the financial year 2023-2024. However, police officials, including Deputy Commissioner David Hudson, emphasize that facial recognition is not the sole determinant for charges. Instead, it functions as a supplementary investigative tool alongside traditional methods.
In addition to Cognitec’s algorithms, the NSW Police use Microsoft’s AI-driven video analytics platform, Insights, for various surveillance functions, including object recognition and vehicle identification, although it does not employ facial recognition. This technology aims to enhance investigation efficiency, yet concerns over its potential bias in high-crime areas have been raised by the AI Review Committee, urging the establishment of guidelines for its data usage in crime prevention.
In conclusion, NSW Police’s cessation of Cognitec’s facial recognition tools highlights growing concerns over their reliability, particularly in accurately identifying individuals from underrepresented backgrounds. The department continues to leverage other technologies, like the NYX system and Microsoft’s video analytics platform, while reinforcing its commitment to employing facial recognition only as an auxiliary investigative resource. Ongoing discussions about the ethical and practical implications of these technologies emphasize the need for balanced usage and vigilant oversight.
Original Source: www.biometricupdate.com
