DoT Develops Facial Recognition Tool for Telecom Operators: ASTR

Context:

The Department of Telecommunications (DoT) has developed an artificial-intelligence-based facial recognition tool that it claims has the capability of running checks on subscriber databases of telecom operators to deduce whether it contains multiple connections associated with the same person.

• The DoT claims the tool — called Artificial Intelligence and Facial Recognition powered Solution for Telecom SIM Subscriber Verification (ASTR) — can potentially bring down cyber frauds by detecting and blocking possible fraudulent mobile connections.

Relevance:

GS III: Science and Technology

Dimensions of the Article:

1. Origins and Testing of ASTR
2. How ASTR Works
3. Actions Taken After Detection

Origins and Testing of ASTR

DoT’s Order for Subscriber Database:

• In 2012, the Department of Telecommunications (DoT) mandated telecom operators to share their subscriber databases, including user pictures, with the department.
• These images form the core database used for the facial recognition algorithm employed by ASTR.

Conceptualization and Design:

• The ASTR project was conceptualized and designed by the DoT’s unit in Haryana between April 2021 and July 2021.
• During this period, the necessary framework and algorithms for facial recognition were developed.

Pilot Project in Haryana’s Mewat Region:

• To assess ASTR’s feasibility, a pilot project was launched in the Mewat region of Haryana.
• Prior to the pilot project, Mewat had approximately 16.69 lakh SIM cards across various telecom operators.
• Through ASTR’s implementation, it was discovered that nearly 5 lakh SIM cards were fraudulent.

Fraud Detection and Validation:

• ASTR’s facial recognition algorithm was employed to analyze the subscriber database and identify potentially fraudulent connections.
• The pilot project in Mewat served as a validation of ASTR’s effectiveness in detecting and identifying fraudulent SIM cards.

How ASTR Works?

Face Encoding:

• ASTR utilizes convolutional neural network (CNN) models to encode human faces in the subscriber images.
• The encoding process accounts for factors like face tilt, angle, opaqueness, and dark colors in the images.

Face Comparison and Grouping:

• ASTR conducts a face comparison for each face against all faces in the subscriber database.
• Similar faces are grouped together under one directory based on their similarities.
• ASTR considers two faces to be identical if they match to a minimum extent of 97.5%.

SIM Detection:

• ASTR detects SIM cards associated with a suspected face within 10 seconds from a database of up to 1 crore (10 million) images.
• It checks if there are more than nine connections linked to a single individual’s photograph, which exceeds the permitted limit set by the DoT.

Fuzzy Logic for Name Matching:

• ASTR employs fuzzy logic to find similarity or approximate matches for subscriber names.
• It can generate related results for a name search, accounting for typographical errors in the subscriber acquisition forms.

Cross-Checking with Different Names:

• ASTR searches the database to identify if the same person has obtained SIM cards using different names.
• This helps detect potential cases of individuals acquiring multiple connections under different identities.

Actions Taken After Detection

Blocking of Connections:

• Once the Department of Telecommunications (DoT) identifies a set of numbers obtained through fraudulent means, it provides a list of these connections to telecom operators.
• Telecom operators then block the identified connections based on the DoT’s list.

Discontinuation of Connections:

• In the initial phase, ASTR analyzed over 87 crore (870 million) mobile connections.
• Using ASTR, more than 40 lakh (4 million) cases of individuals using a single photograph to acquire mobile connections were detected.
• After due verification, telecom operators discontinued over 36 lakh (3.6 million) connections.

Sharing with Other Platforms:

• The list of fraudulent connections is shared with banks, payment wallets, and social media platforms.
• These platforms take action to disengage the identified numbers from their respective platforms.

Collaboration with WhatsApp:

• According to Minister of Communication Ashwini Vaishnaw, WhatsApp has cooperated with the government in disabling accounts created using the identified fraudulent numbers.
• The government is also working with other social media platforms to address similar issues.

-Source: Indian Express