Facial Recognition Technology (FRT)

Why in the News? NITI Aayog's recent release of the ‘White Paper: Responsible AI for All (RAI) on Facial Recognition Technology (FRT)’ is significant. The paper explores FRT as the first use case under the RAI principles, setting out guidelines for the responsible development and deployment of FRT in India.

About Facial Recognition Technology (FRT) Facial Recognition Technology (FRT) is an advanced AI system that identifies or verifies a person’s identity based on facial features extracted from images or video data. The technology uses complex algorithms and models trained on large datasets to recognize and differentiate between faces.

Applications of FRT

• 1:1 Verification:
• Description: Involves comparing a facial map (the unique features of a person’s face) against a specific image in a database to verify identity.
• Examples:
• Smartphone Unlocking: Using facial recognition to unlock mobile devices.
• Access Control: Verifying identities for secure access to buildings or systems.
• 1:2 Identification:
• Description: Involves comparing a facial map against an entire database of faces to identify a person.
• Examples:
• Mass Surveillance: Identifying individuals in public spaces or crowds.
• Law Enforcement: Locating suspects by comparing faces against criminal databases.

Key Features and Technologies

• Algorithm Types: Includes convolutional neural networks (CNNs) and other machine learning models.
• Face Detection: Identifies and locates a face in an image or video.
• Face Recognition: Compares the detected face with stored images to verify or identify the person.
• Feature Extraction: Analyzes unique facial features and creates a biometric template.

Applications and Use-Cases of Facial Recognition Technology (FRT)

Security-Related Uses

• Law and Order Enforcement:
• Identification of Persons of Interest:
• Example: Uttar Pradesh’s ‘Trinetra’ uses real-time facial recognition for identifying suspected criminals.
• Identification of Missing Persons:
• Example: Telangana’s ‘Darpan’ matches photos to identify missing children.
• Monitoring and Surveillance:
• Example: China’s Skynet Project utilizes extensive surveillance for public security and monitoring.
• Immigration and Border Management:
• Example: Canada’s ‘Faces on the Move’ project helps prevent entry using fake identities.
• Crowd Control:
• Example: During Kumbh Mela 2021 in Prayagraj, Uttar Pradesh, Pan Tilt and Zoom Surveillance Cameras managed large crowds and enhanced safety.

Non-Security-Related Uses

• Verification and Authentication:
• Example: Aadhaar Card verification in India utilizes facial recognition for identity verification.
• Example: Contactless onboarding at airports using Digi Yatra simplifies travel procedures.
• Ease of Access to Services:
• Example: The Central Board for Secondary Education’s ‘Face Matching Technology Educational’ facilitates access to academic documents and services.

Risks Associated with Facial Recognition Technology (FRT) Systems

• Inaccuracies:
• Automation Bias and Underrepresentation:
• FRT systems may have biases based on skin tone, race, gender, etc., leading to disparities.
• Technical Factors:
• Intrinsic (e.g., aging, plastic surgery) and extrinsic factors (e.g., lighting, pose variation) can affect accuracy.
• Glitches or Perturbations:
• Minor, imperceptible changes can disrupt system performance.
• Lack of Human Operator Training:
• Mismanagement due to untrained operators handling or verifying FRT outputs.
• Concerns Regarding Accountability, Legal Liability, and Grievance Redressal:
• Complexity of Algorithms:
• Challenges in assigning responsibility for errors or misuse.
• Trade Secrets and IP Protection:
• Protecting the proprietary nature of FRT systems complicates transparency and accountability.
• Rights-Based Issues:
• Purpose Creep:
• Use of biometric data beyond its original intent violates informational autonomy as recognized by the Supreme Court.
• Data Leaks:
• Poor data security practices leading to breaches and unauthorized access.
• Lack of Meaningful Consent:
• Mandatory use of FRT without alternative options undermines consent.
• Private Security Use:
• Excessive surveillance by private firms could lead to unjustified monitoring.

NITI Aayog’s RAI Principles NITI Aayog’s Responsible AI framework outlines principles for the ethical deployment of AI technologies, including FRT. The framework focuses on:

• Transparency: Ensuring clear and understandable processes and decisions made by AI systems.
• Accountability: Holding developers and users responsible for the outcomes and impacts of AI technologies.
• Fairness: Addressing and mitigating biases in AI systems to ensure equitable treatment across all demographics.
• Privacy: Protecting personal data and ensuring its responsible use.

Way Forward: Recommendations of NITI Aayog for Responsible Use of FRT

• Principle of Privacy and Security:
• Data Protection Regime:
• Implement regulations fulfilling legality, reasonability, and proportionality (e.g., Digital Personal Data Protection (DPDP) Act 2023).
• Holistic Governance Framework:
• Define liability for harms or damages caused by FRT systems.
• Adopting Privacy by Design (PBD) Principles:
• Ensure explicit user consent for data collection and usage.
• Principles of Accountability:
• Address transparency, algorithmic accountability, and biases.
• Grievance Redressal System:
• Establish accessible channels for addressing FRT-related issues.
• Ensuring Safety and Reliability:
• Publish standards related to explainability, bias, and error rates in FRT.
• Protection and Reinforcement of Positive Human Values:
• Form an ethical committee to evaluate and oversee the ethical implications and mitigation strategies related to FRT.