Possibilities and pitfalls of facial recognition system | Explained
Facial Recognition Systems (FRS) deployed by Delhi Police during recent student-led protests at Jantar Mantar have come under judicial scrutiny before the Supreme Court of India
A three-judge Bench headed by the Chief Justice of India tagged a writ petition challenging the use of facial recognition and biometric data collection at protests with related matters for detailed examination
Delhi Police told the Court that its deployment was targeted, not indiscriminate — configured to flag only individuals whose photographs already exist in police records (wanted persons, absconders, habitual offenders, history-sheeters), and did not profile every person present
Petitioners allege biometric data was collected without consent and hosted by private facial-recognition vendors, in violation of data-protection norms
The episode has revived a long-running debate on the absence of a dedicated legal framework governing facial recognition technology (FRT) in India
How Facial Recognition Technology Works
FRS uses computer-vision algorithms to detect a face in an image or video feed, extract distinctive facial landmarks (a "facial signature" or template based on features like eye spacing, jaw contour, nose shape), and match that template against a reference database. Two functional modes are typically distinguished: 1:1 verification (confirming a person is who they claim to be, e.g., unlocking a phone) and 1:many identification (searching a face against a large database to establish identity, as in police deployments).
Key Details
- Core pipeline: face detection → feature extraction (template generation) → matching against a reference database → confidence score output
- Accuracy is highly dependent on image quality, lighting, angle, and the diversity of the training dataset; documented biases include higher error rates for women and darker-skinned individuals in several global studies
- Police deployments in India have shown low field accuracy — Delhi Police pilots have been reported with accuracy well below claimed benchmarks in real-world (as opposed to lab) conditions
The Supreme Court's scrutiny centres on exactly this identification-mode use — matching faces captured at a protest against police watchlists — where both technical reliability and the absence of consent are contested.
Right to Privacy — K.S. Puttaswamy v. Union of India (2017)
A nine-judge Bench of the Supreme Court unanimously held in K.S. Puttaswamy v. Union of India (2017) that the right to privacy is a fundamental right, protected as an intrinsic part of the right to life and personal liberty under Article 21, and also flowing from the freedoms guaranteed under Part III of the Constitution. The judgment laid down a three-fold test for any state action restricting privacy: legality (existence of a law), legitimate state aim, and proportionality.
Key Details
- Decided by a 9-judge Constitution Bench, 2017; overruled earlier contrary observations in M.P. Sharma (1954) and Kharak Singh (1962)
- Explicitly flagged biometric data (including facial data) as a sensitive category of personal information
- The proportionality test from Puttaswamy is the standard against which mass-surveillance tools like FRS are now tested in Indian courts
The absence of a specific statute authorising police FRS deployment means such use is tested directly against the Puttaswamy proportionality framework — the core constitutional question now before the Supreme Court.
Regulatory Gap: DPDP Act, 2023 and the Missing FRT Law
The Digital Personal Data Protection (DPDP) Act, 2023 is India's first comprehensive data-protection statute, governing the processing of digital personal data on principles of consent, purpose limitation and data minimisation. However, it was not designed specifically for law-enforcement biometric surveillance, and it carries broad exemptions for state functions relating to law and order and public interest, leaving police use of FRS in a regulatory grey zone.
Key Details
- DPDP Act, 2023 passed by Parliament August 2023; treats biometric identifiers as personal data requiring lawful basis for processing
- Section 17 of the DPDP Act exempts processing by the State for functions like prevention/detection of offences, subject to conditions — a key point of contention for FRS use by police
- The National Automated Facial Recognition System (NAFRS), for which the National Crime Records Bureau (NCRB) issued its first RFP in 2019, remains without a dedicated parliamentary statute governing its operation
- A private member's Facial Recognition Technology Regulation Bill has been proposed but not enacted
Petitioners' core legal argument is that facial data was collected and stored (including by private vendors) without meeting DPDP Act consent/purpose-limitation standards, while police rely on the Act's law-enforcement exemption — a gap the Supreme Court is now being asked to resolve.
- K.S. Puttaswamy v. Union of India: decided 2017, 9-judge Bench, privacy read into Article 21
- DPDP Act enacted: 2023; core principles — consent, purpose limitation, data minimisation
- NCRB's first National Automated Facial Recognition System (NAFRS) RFP: issued 2019
- Delhi Police FRS deployment (recent protests): stated to match only against existing criminal records, not all attendees
- Documented field accuracy issue: Delhi Police FRS pilots historically reported with error rates far above vendor claims, particularly in identifying women and children