'AI Smart Policing' Making Ordinary Detectives Great Detectives Emerges [Reading Science]

[Asia Economy Reporter Kim Bong-su] The protagonist of British author Arthur Conan Doyle's classic detective novel, 'Sherlock Holmes,' boasts 'computer-like investigative skills' that do not miss even the smallest clues and data. He can determine whether someone is a criminal just by the color of the dirt on their shoes. Now, with the advancement of scientific technologies such as artificial intelligence (AI), robots, and big data, ordinary police officers and investigators can become real-life 'Sherlock Holmes.' Beyond simply developing advanced security equipment, smart policing is increasingly active in predicting and preventing crimes in advance and swiftly apprehending criminals.

◇ What is Smart Policing?

According to the Ministry of Science and ICT, smart policing is generally defined as 'police activities that utilize science and technology together with the community to maintain and preserve the peace and order of the nation and society.' Specifically, it refers to policing activities that operate limited police resources selectively and intensively by utilizing various scientific technologies such as artificial intelligence, big data, drones, and life sciences in response to rapidly changing security environments, including the surge of new and intelligent crimes. The 'SMART' in smart policing stands for ▲Strategic Management ▲Analysis & Research ▲Technology. Furthermore, collaboration among police, research institutions, and local communities for technology development and solution enhancement is a core element of smart policing. Smart policing technologies are developed and advanced through processes where police, research institutions, and communities share crime data, identify problems, and resolve them.

◇ Increasingly Intelligent Crimes

According to the National Police Agency's crime statistics, in 2019, the incidence of intelligent crimes was higher than that of traffic (727.8), violence (555.3), and theft (360.6) crimes. The occurrence rate per population was the highest at 735.8. Cybercrime is also becoming increasingly rampant. The global cybercrime damage scale was $3.5 billion in 2019 but surged by 20% to $4.2 billion in 2020. The number of victims also increased significantly from 467,000 to 791,000 during the same period. Accordingly, smart policing to address such intelligent and cyber crimes is becoming more active. The global public safety and security market has grown at an average annual rate of 7.2% over the past five years. It is expected to increase from $365 billion in 2020 to $516.5 billion by 2025.

[Image source=Yonhap News]

◇ Catching Criminals with Data

Technologies that analyze crime status using data and identify concentrated occurrence areas for preemptive prevention are advancing day by day. After the 9/11 incident, the New York Police Department collaborated with Microsoft to build a system that automatically detects or predicts crime scenes in advance based on big data such as photos, videos, and audio. It tracks surveillance targets using more than 6,000 cameras. It can also detect radiation, nuclear weapons, and provides data such as emergency calls, suspect arrest records, and suspect vehicles. Similar services are widespread. In China, services such as vehicle license plate recognition, criminal facial recognition, and crime risk behavior detection have been introduced by analyzing video data from cameras. An Israeli company also collects real-time data using cameras and UAVs and provides services that improve urban safety and security situations through behavior analysis, facial recognition, and risk situation identification based on this data. Additionally, there are solutions that collect and analyze gunshot sound data using gunshot detection sensors and GPS devices to help police respond quickly to the scene.

In particular, systems that predict the likelihood of crimes and potential victims through AI analysis to prevent crimes in advance are being established one after another. PredPol, introduced by the Los Angeles Police Department, is a representative example. It analyzes past crime data such as crime type, location, date, and time using AI to predict future crime types and locations. During the first six months of PredPol's application, the frequency of robberies reportedly decreased by 19%, showing significant effects. In Korea, the National Police Agency is building the 'CLUE' system, which analyzes crime data and public data with AI to provide results such as crime trends, similar case recommendations, and crime predictions.

◇ Finding Missing Persons Made Easier

Technologies that help easily find missing persons based on advanced AI using information such as movement routes, composite sketches, and locations are also being researched and developed. Recently, the Korea Institute of Science and Technology (KIST) developed a technology that predicts composite sketches of long-term missing persons by analyzing age-related characteristic points?such as facial shape changes, wrinkles, pigmentation, and skin color changes?using AI on 10,000 Korean composite sketches including age information. Additionally, a research team at Johannes Kepler University in Austria created an autonomous flying drone equipped with thermal sensors and applied aerial optical section algorithms and AI technology to search for missing persons in mountains or large forests.

◇ AI Preventing Cybercrime

AI technology development to respond to increasingly sophisticated and intelligent cybercrimes is also active. Recently, KB Kookmin Bank developed an electronic financial fraud detection technology by reproducing fraud methods with white-hat hackers and anomaly transaction detection system managers over about six months and analyzing patterns with AI technology to counter phishing scams. It blocked about 170 fraudulent transfers and protected approximately 2.5 billion KRW of customer assets. In the U.S., Nasdaq is cooperating with AI company Digital Reasoning to develop AI technology that analyzes language used in messengers and emails to crack down on insider trading and stock price manipulation. Real-time obscene image filtering systems have also become more sophisticated. Naver developed and is operating 'X-eye,' an obscene content filtering AI system that classifies about 4 million images by type and, after AI training, achieves 98.1% accuracy in blocking exposure of obscene images and videos.

◇ AI Apprehends Criminals

The U.S. Federal Bureau of Investigation (FBI) has introduced AI-based technology that compares police mugshot databases with faces in social media and CCTV images for criminal investigations. The UK's Serious Crime Directorate uses AI-based investigative robots to review, classify, and summarize about 600,000 pages of documents daily, dramatically shortening investigation periods. A victim statement preparation system has also emerged. Selvas AI developed software that uses voice recognition technology to automatically convert conversations between investigators and victims conducted in front of a dedicated microphone into statement documents. The AI automatically provides additional question lists and Supreme Court precedents based on the victim's testimony. In Korea, the National Intelligence Resources Service and the National Police Agency developed a system that learns text data such as location, time, and crime methods with AI to find whether suspects have committed additional crimes in the past.