National Crime Records Bureau (NCRB) under the Government of India has released a tender asking for bidders to help create Automated Facial Recognition System (AFRS). The objective is to leverage the power of facial recognition technology to make the security forces more efficient.
The technology system which has been proposed will make an extensive database of photos belonging to Indian citizens using which machine learning models will be trained. The criminals will be identified using CCTV footage from the database, verified, and information will be distributed to all law enforcement agencies in real-time across the country.
Experts have touted that using facial recognition technology to identify and solve crime may be one of the best applications of the technology. At the same time, others have expressed concerns that facial recognition may indirectly make India a surveillance state, following the footsteps of our neighbouring country China. The concerns are also based on the lack of proper data security standards, absence of law in privacy and data governance on the federal level.
The Mechanism Of AFRS
As part of the Automated Facial Recognition System (AFRS), images from various sources like CCTV, phone cameras, sketches, social media and others could be instantly matched with the photo database. According to the circular, this would help identify criminals, missing children, unidentified deceased bodies and other aliens persons on the borders. The end product will be made available on the web platforms and mobile devices be in Android, iOS or Windows.
NCRB is also working for the integration of fingerprint data under the NAFIS program to be integrated with Crime and Criminal Tracking Network & Systems (CCTNS). Here, facial data that includes all metrics of the face, combined with fingerprint, will help the law enforcement agencies in their investigations. The proposal for mobile data terminals to each police station has been made under CCTNS Phase-II. The facial image database will also be integrated with various other systems & databases like ICJS, IVFRT, ICJS, Immigration Visa Foreigner Registration Tracking (IVFRT) and with all the police systems in different states.
Conditions Specified By NCRB For Private Vendors For Automated Facial Recognition System
There will be a bid process to choose the right vendor for the technology, according to the published circular. As part of the conditions, the system will have to offer logical algorithms and user-friendly, simple graphical user interface. At the same time, it will depend on the state-of-the-art machine learning capabilities to achieve its objective of generating accurate alerts by analysing the massive database.
NCRB also specified that the system should be able to work despite changes in facial expression, lighting, age, hairstyle, facial hair, scars and tattoos that may occur in facial data over time. Not just that, the technology should also have the capability to identify a facial image by automatically extracting it from solid backgrounds and other landmarks. The system will be designed where algorithms, modes and search depths work for varying environments like fast-moving situations and supporting not just CCTV but also infrared and thermal cameras. It seems the specified conditions for the system will require state of the art solutions from the vendor.
Challenges That Come With Facial Recognition Technology
The facial recognition technology depends on how well the system is trained on the datasets and barring China which with the largest facial photo database, no other country has achieved high precision in mass deployments. Many experiments have also proven how facial recognition systems may fail to identify objects accurately, as highlighted by multiple research works. Facial recognition software tested by the UK’s Metropolitan Police generated false positives in more than 98% of match alerts.
In another experiment, in Germany, researchers found the accuracy of facial recognition technology between 17-29% and full of false positives. There are also concerns around demographic bias may unintentionally creep into the system during training the ML models. According to an MIT study, facial recognition systems make more significant errors in recognising subjects with darker skin complexion and females.
Facial Recognition System: Data Security May Be The Biggest Roadblock Here
The notice has identified the related security challenges on unauthorised access to database or application, accidental modifications or deletions, unintentional modifications or deletions, encryption/decryption of records while storing/loading the records to/from the database. It has been proposed that AFRS needs implementation of an access control policy, MD 5/SHA encryption layer, prevention from BRUTE Force Attack, SQL Injection, other vulnerability patches, etc.
Considering the advanced state of state-sponsored hackers, it will remain uncertain how government agencies help protect facial images databases. According to reports, the entire Aadhaar database had been breached in the past, leaking data to hackers on the dark web. A new database containing extensive facial data belonging to Indian citizens may become another juicy target for hackers across the globe.
Privacy Vs Law & Order: What Will Be The Priority?
Looking at the positive side, the use of facial data in criminal investigations will likely bring down crime rates. To achieve this, it is expected that the government will be working on a nationwide installation of closed-circuit cameras that can capture facial imagery in real-time. Similar systems have found success in the US and China, where criminal data is shared across different departments, helping to solve investigations effectively.
But in a country like China, facial data is not just used to catch criminals but also to monitor citizens and create social credit scores. To prevent such a scenario, it may be imperative for the government to clarify its data privacy law and have a governance body which ensures data is used ethically and for helping the nation in the right direction.
The concerns for privacy is highlighted by a recent instance where San Francisco – the technology hub of the world made it illegal to use facial recognition collected by police forces. The reason cited is the lack of proper regulation on how such personal data will be managed and secured. Given the lack of privacy laws, experts say the proposed AFRS may pose a great challenge in determining the accountability of any unintentional misuse of facial data. This particularly is of greater importance in India, a country which guarantees the right to individual privacy as part of the constitution.
Enjoyed this story? Join our Telegram group. And be part of an engaging community.
Provide your comments below
What's Your Reaction?
Vishal Chawla is a senior technology journalist at Analytics India Magazine, and writes on the latest trends in analytics, AI and other digital technologies. Previously, he was a senior correspondent for IDG ComputerWorld and CIO. Vishal can be reached at email@example.com