Compliance with Indian law is starting to place a lot of importance on facial recognition technology. Delhi police, which investigated the identification of people involved in civil unrest in northern India in recent years, said they would consider an accuracy of 80 percent or higher to be a “positive” match, according to documents obtained by the Internet Freedom Foundation through a public registry. request
The arrival of facial recognition in India’s capital region marks the expansion of India’s law enforcement officials using facial recognition data as evidence for possible prosecution, raising alarms among the privacy and civil liberties experts. There are also concerns about the 80 percent accuracy threshold, which critics say is arbitrary and too low, given the potential consequences for those flagged as a match. The lack of a comprehensive data protection law in India makes matters even more worrying.
The documents further state that even if a match is less than 80 percent, it would be considered a “false positive” rather than a negative, making the individual “subject to due verification with other corroborating evidence”.
“This means that even if facial recognition doesn’t give them the result that they themselves have decided is the threshold, they will continue to investigate,” says Anushka Jain, the IFF’s associate policy adviser for surveillance and technology, who request this information. . “That could lead to harassment of the individual just because the technology says they look like the person the police are looking for.” He added that this move by the Delhi Police could also lead to harassment of people from communities that have historically been targeted by law enforcement officials.
In response to the IFF’s records request, the police said they are using convict photographs and case file photographs to run facial recognition. They added that they could be used as evidence, but declined to share more details. They clarified, however, that in case of a positive match, police officials would conduct additional “empirical investigation” before taking any kind of legal action. Delhi Police did not respond to WIRED’s emailed requests for comment.
Divij Joshi, who has spent time researching the legality of facial recognition systems, says the threshold of an 80 percent match is practically meaningless. Joshi explains that accuracy numbers are highly dependent on the conditions for testing models of facial recognition technology against specific benchmark datasets.
“Typical accuracy with facial recognition or machine learning systems is determined by comparing a model developed on training data and validation data to a reference data set,” says Joshi, a PhD student at University College London. “Once the training data is modified, it should be compared against a third-party dataset or a slightly different dataset.” This comparison, he says, is what is typically used to calculate percent predictive accuracy.
Evidence of racial bias in facial recognition models makes using the technology problematic. And while many variables affect the accuracy of facial recognition systems, widespread police use of a system with an 80% accuracy threshold appears to be highly unusual. A 2021 study by the US National Institute of Standards and Technology found that systems used to match a single scan of travelers’ faces with a database containing their photos had an accuracy of 99.5 percent or better. Other studies, however, have found error rates as high as 34.7 percent when used to identify women with darker skin.