Indian Prime Minister Narendra Modi attended an Indian people’s rally in Ramlila Maidan, New Delhi, on December 22 , the first time New Delhi police have used facial recognition software to screen people, Indian media reported. It is also the first time New Delhi police have used a set of facial images collected from videos taken during various protests in the city to screen “security suspects” at the prime minister’s rally.
New Delhi police in March 2018 purchased automatic facial recognition system (AFRS) software that identifies missing boys and girls by identifying photos, following an order from the High Court of New Delhi concerning a case involving a missing child.
Prior to 22 December, AFRS had used only three – twice in the Independence Day parade and once in the Republic Day parade.
New Delhi police have so far created more than 150,000 “historical” photo sets for routine crime investigations, the sources said. Another data set used to monitor sensitive public events contains more than 2,000 images of terror suspects, as well as newly added images of so-called “harassers and outlaws.” The use of AFRS by the police is significant in view of the ongoing protests throughout the city against the new Citizenship Amendment Act.
Police in New Delhi have long been filming major protests in the city. Now they’re entering the footage into AFRS, which extracts the protesters’ “recognizable faces” into their datasets. Once extracted, the sources said, the data would be manually screened to identify and retain “habitual protesters” and “outlaws.”
For the first time, the “selected protesters” data set was used to keep out “criminals who may have held up signs or slogans,” the sources said. “Each participant at the rally was caught on camera at the door of the metal detector and matched the live video to the facial data set in the control room set up at the site within five seconds,” explained a police official involved in the security program.
“Any FRS (facial recognition research software) will remain substantial only if the relevant database is qualitative and quantitative,” a New Delhi police spokesman said in an email to the Indian Express. The initial focus was on security-related aspects of the Independence Day and Republic Day arrangements, which were achieved through the establishment of the latest data sets for terrorist suspects, etc. “
“In the next phase, we will also focus on law and order and, accordingly, extend the data set to those with known criminal records in the relevant categories, as well as to laws and suspects identified through a large number of archived video and behavioral analysis conducted in sensitive public protest sites.” We used these data sets on Last Sunday based on reliable intelligence on suspicious interruptions. “
Asked what checks and balances were in place to avoid image abuse and privacy, the spokesman said: “The New Delhi police did not go into details, but ensured that the best industry standard checks and balances were in place to prevent possible data abuse.” Such datasets are not permanent and are modified periodically. When these data sets were established, racial or religious characteristics were never a related parameter. “
Currently, the AFRS version of the Delhi police can handle different facial data sets, with a maximum capacity of 300,000 photos at a time. One of the sponsors of innefu Lab in New Delhi provided the system to the police.