Growth in the Use of AI Based Surveillance Technologies

Authored by Debashis Guha Associate Professor and Director – Machine Learning

Government surveillance is one of the fastest growing applications of Artificial Intelligence and related technologies. A recent article from the Carnegie Endowment for International Peace reports that AI-based surveillance is already used in 75 out of 176 countries in the world. They report that these tools come from three major families of usage, smart/safe city face platforms, face recognition, and smart policing.

A “Smart City” or “safe city” is a platform that facilitates service delivery, city management, and public safety. They incorporate sensors, facial recognition cameras, and police body cameras that are networked together and connected to “intelligent command centres”. The command centres run algorithms to detect anomalous movements, weapons, and crowds, often using deep neural networks. The algorithms are designed to prevent crime, ensure public safety, and respond to emergencies. One of the principal vendors of safe city platform software is the Chinese company Huawei, the world’s leading maker of surveillance tools. Their safe city systems have been deployed in the cities of Marseille and Valenciennes in France, in addition to several Chinese cities.

Another major tool used in AI based surveillance is face recognition. Its most basic function is to match live data to tagged samples stored in a database, using a set of biometric features.  Other possible uses are to detect sentiment displayed by a collection of faces in a crowd, or some other collective feature. Face recognition methods can be used to detect wanted criminals in a crowd, as has been donein China, or to monitorand arrest rioters, as was done in Baltimore, USA. A toolreleased by NEC uses face recognition along with gesture recognition, and is designed to detect shoplifting at retail establishment. A wide assortment of surveillance systems using face recognition and other video and image detection methods have been deployed along the US Mexico border to detect illegal entry.

Smart policing uses data-based analytics to facilitate investigations and police response, along with algorithms that make predictions about future crimes. Smart policing can take many forms. Autonomous drones can be used to capture video and images that are scanned by an AI system to detect or anticipate anomalous events. Such a system, consisting of drone-based streaming video, and predictive video analytics software has been used in several locations in India, as was reported as far back as in 2016. Other smart policing tools involve the use of predictive analytics to anticipate criminal acts by gangs, and these have been used in California.

China is the most important source for surveillance technologies for much of the world. Some of the leading Chinese companies in this space are Huawei, Hikvision, Dahua, ZTE and. Some of the other leading companies are NEC of Japan, and IBM, Palantir, and Cisco of USA.

Researchers at the Carnegie Endowment have collected data across the world on the use AI technology for smart/safe city, facial recognition and smart policing. An extract from this global table is presented below.

Country	Safe City	Face Recognition	Smart Policing
Brazil	✔	✔	✔
China	✔	✔	✔
Egypt		✔	✔
France	✔	✔	✔
Germany	✔	✔	✔
Ghana	✔
India	✔	✔	✔
Japan	✔	✔	✔
Nigeria	✔	✔
Russia	✔	✔	✔
South Africa	✔	✔	✔
Switzerland	✔	✔
UK	✔	✔	✔
USA	✔	✔	✔

As the table above shows, India uses all three major families of tools for carrying out AI based surveillance. Recently, the Government of India has launched a mission to expand the scope of its surveillance activity in a big way. Bloomberg reports that

Prime Minister Narendra Modi’s government will open bids next month to build a system to centralize facial recognition data captured through surveillance cameras across India. It would link up with databases containing records for everything from passports to fingerprints to help India’s depleted police force identify criminals, missing persons and dead bodies.

The link to the tender details shows that a Request for Proposal for a National Automated Facial Recognition System (NAFRS) was published on 28-Jun. This was a detailed document that ran to 172 pages, and this has now been followed by an open bid for the delivery of this automated system. In addition, there was an expression of interest for the “Selection of System Integrator for Supply, Installation, Maintenance and operation of Tools to Identify Child Pornography/Obscene Contents in the Online Space”, which will be an online surveillance tool.

The detailed RFP for the NAFRS shows that it will connect CCTV cameras from across the nation and combine this data with the biometric and other data from the tax, passport, and UIDAI databases. The resulting database will probably be the most voluminous store of personal data anywhere in the world. Since the accuracy of deep learning and most other AI systems improves with dataset size, this will be a golden opportunity to build the most accurate automated learning systems for image and video analytics and face recognition, and also for developing the world’s best public safety and smart policing platforms.