Sometime early last month, a ‘Bharat Jodo (Unite India)’ rally was organised at the Jantar Mantar protest site in Delhi with members of several right-wing organisations in attendance. The ostensibly minor event made to the headlines the next day after several videos in which people could be heard allegedly raising inflammatory anti-Muslim slogans went viral.
Under intense pressure to arrest and identify the accused, the Delhi police lodged a case and turned to their Facial Recognition System (FRS) for assistance.
A few videos of the event and a poster with faces of eight persons were fed into the FRS. “The software was able to identify two persons from the poster via the driving licence database we have access to, but they have not been arrested yet,” said an officer privy to the case.
This is the third time in two years that the police have used FRS during a major investigation. The software was used to probe a 2020 north-east Delhi communal riots case, and the January 26, 2021, Red Fort incident, when farmers stormed the monument during a tractor rally against new agriculture laws. Both incidents were widely covered by TV news channels and countless videos were shared on social media.
The police currently use a “basic” level of FRS, which does not function in real-time. The force has access to all CCTVs in the city but is hamstrung by a lack of access to metadata: it can seemingly only tap into the driving licence and criminal databases of some adjoining States. Photos of missing and found children, and missing persons, have also been fed into the system. Sources said the Election Commission, UIDAI and other Ministries do not share data with the force.
Several limitations
While probing the death of a Hindu man during the riots, the Investigating Officer (IO) sent notices to TV news channels to share videos of the incident. The IO also wrote to the Delhi government requesting it to share relevant CCTV footage. The request had to be hand-delivered. “At the time, getting CCTV footage was a task. But now the situation has changed, they [the authorities] have given us a password and username for the CCTVs and we can monitor footage in our areas,” the officer said.
The police zeroed in on a suspect after examining all the footage. “The video and a mugshot of the suspect was uploaded to the FRS, which returned four matches. One of them was a Muslim; he turned out to be innocent. We did not make any arrest using FRS in that case,” the officer said.
The software examines photos using parameters such as facial features, age, skin colour, etc. “It gives a 70-80% match against the photo put in the system. It only works as an aid. If it filters one lakh photos down to 50, then it helps in policing,” said a senior police officer.
The software has had its successes. The officer said that the FRS is frequently used to trace missing children and identify bodies. When a child is reported missing, the IO is expected to upload details along with a photo on ZIPNET (Zonal Integrated Police Network). When a child is found, a photo is uploaded and the software sifts through the database for any matches.
All Delhi police stations have access to the FRS and a few officers in every police station have been given the username and password for it, the officer said.
Restricted access
While the force is still coming to terms with how the new software can be utilised, debates rage over its misuse and privacy concerns. However, t he technology available to the police is a far cry from what sceptics fear — total or real-time surveillance.
A sub-insp ector said the police station he is posted in has no access to FRS as far as he is aware. “On the ground, most officers do not even upload the details of missing or found children. When they are not using ZIPNET properly, how will they get around to using FRS?” he said.
An officer in another police station said they do not have ready access to the software. In the rare case when FRS is required, permission is needed from senior officers. “We have to file an application along with FIR number and then send it to the Crime Review Office in Kamla Market police station. It is such a lengthy process that I am not sure if it is ever used in daily investigations,” the officer said.
Another officer said FRS is being used in the e-beat book application, which was launched recently. Under the initiative, beat officers are given a smartphone that has facial recognition feature along with access to criminal database. Whenever a policeman spots someone suspicious, they can take a photo and upload it to the application, and find out whether the person has a criminal history.
An officer who investigated the 2020 riots said the software is not very efficient as it does not consider the age parameter. “If the person’s driving licence is 20 years old, then it is impossible that the FRS will be able to find that person,” the officer said.
Software upgrade
The so ftware was bought by the Delhi police in 2018 from Innefu, co-f ounded by Tarun Wig.
Speaking toThe Hindu, Mr. Wig said Delhi Police procured the software to find missing children and that he was not aware of the agency using it to find suspects.
On privacy concerns regarding the FRS, Mr. Wig said, “Delhi Police has around one lakh personnel and the city’s population is around 1.52 crore. They cannot really monitor so many people. That is where predictive policing comes into play. Its uses and misuses exist but that can be said about any technology,” he said, adding that they are regularly upgrading the software to increase its efficiency and potential.
The prospects of this getting used widely, more efficiently and hence more intrusively thus remain.
