India is beefing up police force with AI tools to nab criminals

Science fiction has always had an uncanny knack of predicting future societies. In Philip K Dick’s 1956 short story Minority Report, we encounter a police department of the future preventing crimes before they actually happen with the help of computers and three mutants. 

While we haven’t got hold of mutants as yet, fighting crime with the help of super intelligent computers is almost a reality now. Today, India’s top police forces are deploying artificial intelligence (AI) to detect and prevent crime which brings things somewhat close to Dick’s vision of the future.

In 2009, the United Progressive Alliance (UPA) government launched the ‘Smart Police’ reforms, under which a Crime and Criminal Tracking Network and System (CCTNS) was set up, connecting 15,000 police stations and 5000 supervisory police officers across the country. Since then law enforcement agencies have built networks of closed-circuit television (CCTV) for digital traffic violation detection. These are now being scaled up further on the digital and technological frontier.

“Four state governments are introducing AI in their police operations,” says a cyber-security partner working with a multinational services organisation. “However, the reach of the technology and its scalability is still about three to five years away,” he adds.

States such as Kerala, Maharashtra and Andhra Pradesh have begun using Computer Emergency Response Teams (CERT.In) to source, design and implement AI solutions into their police forces. CERT-In is the nodal agency under the ministry of electronics and information technology, which  monitors cyber security-related crimes.

One of the earliest instances of the use of AI for crime mapping and predictive analytics took place in New Delhi. In 2017, the state police implemented a software called CMAPS (Crime Mapping Analytics and Predictive System) which uses the Indian Space Research Organisation’s satellite images and spatially locates suspicious telephonic conversations across the national capital region.

“We are applying analytics to data collected across the national capital region. The software records calls made to the PCR number (#100), First Information Reports, and also spot reports made either by local police officials, patrolmen or the PCR,” says a senior official with the Delhi Police.

The software indexes each crime and plots the pattern across the city, enabling the police to look for patterns of criminal activity based on spatial and/or temporal factors.

“As the live-map displays ‘hot-spots’ of crime, it helps in improving our deployment abilities and patrolling strategies in terms of what to focus on and where and when,” the Delhi Police official says, adding that this can help the police prevent crimes.

While the government is pushing its agencies to modernise the crime-solving abilities of local police forces, the effort is being aided by private entities as well. For example, Staqu, an AI development company, is working with around nine police departments across the country to digitalise the police records and provide them with meaningful insights using software algorithms. In association with the Alwar police in Rajasthan, the company has launched an app called ABHED (AI Based Human Efface Detection) which allows cops to register, identify, track and monitor criminals with the help of mobile phones.

“State-level police authorities already had a data base of criminals. We are digitising these old files of criminal records,” says Atul Rai, chief executive officer of Staqu. The company is also working with Punjab police’s Organised Crime Control Unit, the Uttarakhand state police, as well as with a few other intelligence agencies across the country.

State governments see the importance of digitising criminal records and churning the information to come out with predictive models. For example, the CERT team in Maharashtra is implementing a predictive policing technology for the Mumbai police. Under this, all the new data being collected, including the biometric details of the suspects, are fed into a database to build an Automated Multi-Modal Biometric Identification System. PricewaterhouseCoopers has been roped in as a consultant to the project.

“This initiative will help the police force triangulate a person and get a 360 degree view based on multiple factors,” the cyber security professional cited above says.

Balsingh Rajput, superintendent of police (SP), Cyber-Maharashtra, who heads CERT-Maharashtra told Business Standard in an earlier interview that suspects will now have their fingerprints, iris and face scanned. The data will then be entered into the system and matched to any existing records.

So far, the system has been introduced in 13,200 police stations across Maharashtra, though its hardware component is yet to be delivered to half of them. “The idea is to predict crime, detect patterns of the crimes and probable geographical areas where they may occur,” Rajput says.

The ultimate goal is to be able to quickly triangulate possible suspects and to develop real-time surveillance models so as to track them in the physical as well as in the virtual world. However, for scalability, efficiency and effectiveness, the system will have to be integrated with all databases — residential, criminal and perhaps financial — say experts. The tracking will also involve data scraped from social networks which is often highly personal. In a recent discussion paper, the NITI Aayog too says that monitoring of social media can ensure public safety.

But as AI or predictive policing tools like CMAPS are implemented across the CCTNS or scaled across all police stations in the country, there are legitimate concerns over data security and individual privacy.

The National Crime Records Bureau said last month that it would like to have ‘limited access’ to Aadhaar data, which raises fears of a police-surveillance state. However, Rajput of CERT-Maharashtra says that the predictive crime software that the state police is building is not connected to UIDAI and will not request Aadhaar data to trace criminals.

Rai of Staqu also says that the programmes developed by the company do not give access to every police personnel, but only to a select few based on “three levels of authentication through OTPs and IP-based security authentication.”

Of course, we are still a long way from tapping the full potential of predictive AI in fighting crime. The question, though, is how far we should go in using these tools. At the end of Minority Report, the flaws in the mutants and the computer system leads to the entire programme being scrapped over fears of thousands of people being tried for crimes they would have never committed.
 

And truth, after all, can turn out to be stranger — and more dystopic — than fiction.

Smart Policing

A Crime and Criminal Tracking Network and System connects 15,000 police stations and 5,000 police officers across the country

AI is used for crime mapping and predictive analytics in Delhi

A software uses ISRO satellite images and spatially locates suspicious telephonic conversations across NCR

The data is used to plot the crime pattern across the city