Earlier today, Home Minister Amit Shah stated that facial recognition technology is being used extensively to identify and arrest perpetrators of violence and riots that have shaken up Delhi over the past few weeks. Speaking at the Lok Sabha, he said, “We are using face recognition software to identify people behind the violence. We have also fed Aadhaar and driving license data into this software, which has identified 1,100 people. Out of these, 300 people came from UP to carry out violence.”
The technology that India’s Home Minister is referring to is known as Advanced Facial Recognition Software (AFRS), which was acquired by Delhi Police back in 2018 in an exercise that aimed to identify missing children and catch criminals involved in child trafficking and other related acts. The tool was developed by INNEFU Labs, an Indian startup that works on artificial intelligence tools, and built the machine learning infrastructure that is deployed by the Delhi Police.
As its founder, Tarun Wig, explained in an interview with News18, “The original database for the images depend on what the client feeds our tool. This is under the discretion of the customer, and if they want, they can even take data from Google, Facebook and other public sources, and ingest it into the system to recognise the faces. We do not play a role in the implementation of the software and their uses, for any customer. The same holds true for Delhi Police as well.”
It is this flexibility that is reportedly being used by the Delhi Police to identify individuals involved in the recent Delhi riots, according to Shah. Speaking at the Lok Sabha, he added that 40 teams have been formed within Delhi Police to track down the said 1,100 individuals accused of spreading violence during this period. He further said, “We entered Voter ID data and other information available with the government to identify people who were involved in the riots. Arrests will be made only in cases where there is hard evidence against a person.”
Explaining the working of his AI face recognition model to News18, Wig said, “What AFRS does is detect and extract the faces out of an image. Every face is then converted to a vector of 512 values. Then, the software calculates the shortest distance between two vectors in a chosen database, and the closest matches are typically the accurate recognition of a face.” This chosen database is what Home Minister Amit Shah referred to at Lok Sabha today, which has reportedly helped in the identification of 1,100 perpetrators already.
Going forward, it remains to be seen how many arrests are made based on AFRS’ results. “It is a software, so it does not differentiate on the basis of religion,” added Shah on the matter.