Crime is down but it is changing, said Rt Hon Theresa May MP Home Secretary, UK. While traditional high volume crimes like burglary and street violence have more than halved, previously ‘hidden’ crimes like child sexual abuse, rape and domestic violence have all become more visible, if not more frequent, and there is growing evidence of the scale of online fraud and cyber crime.
As with so many of the challenges we face as a society, the prevention of crime is better than cure. Stopping crime before it happens, and preventing the harm caused to victims, must be preferable to picking up the pieces afterwards.
Data and data analytics, tools have become critical in successfully preventing crime. Many police forces are already trialling forms of ‘predictive policing’, largely to forecast where there is a high risk of ‘traditional’ crimes like burglary happening, and plan officers’ patrol patterns accordingly, says UK’s Modern Crime Prevention Strategy. Data analytics can be used to identify vulnerable people, and to ensure potential victims are identified quickly and consistently.
China, a surveillance state where authorities have unchecked access to citizens’ histories, is developing artificial intelligence based tools that they say will help them identify and apprehend suspects before criminal acts are committed.
China planning to use AI technology to predict and prevent crime
China’s crime-prediction technology relies on several AI techniques, including facial recognition and gait analysis, to identify people from surveillance footage, according to The Financial Times. In addition, “crowd analysis” can be used to detect “suspicious” patterns of behaviour in crowds, for example to single out thieves from normal passengers at a train stations.
Facial recognition company Cloud Walk has been trialling a system that uses data on individuals’ movements and behaviour — for instance visits to shops where weapons are sold — to assess their chances of committing a crime. Its software warns police when a citizen’s crime risk becomes dangerously high, allowing the police to intervene.
“If we use our smart systems and smart facilities well, we can know beforehand . . . who might be a terrorist, who might do something bad,” said Li Meng, vice-minister of science and technology.
Another example of AI use in Chinese crime prediction is “personal re-identification” — matching someone’s identity even if spotted in different places wearing different clothes, a relatively recent technological achievement.
“We can use re-ID to find people who look suspicious by walking back and forth in the same area, or who are wearing masks,” said Leng Biao, professor of bodily recognition at the Beijing University of Aeronautics and Astronautics. “With re-ID, it’s also possible to reassemble someone’s trail across a large area.”
Durham Constabulary Deploy AI for Crime Prevention
Durham Constabulary is preparing to trial an artificially intelligent system to help officers decided whether or not to keep a suspect in custody.
The Force, will use the Harm Assessment Risk Tool (Hart) to help officers decide if a suspect can be released from detention, based on the probability of offending once released. Hart has been trained on five years of the Force’s data (from 2008 – 2012), and will classify a suspect as either low, medium, or high risk of offending. The system was tested from 2013, with forecasts that a suspect was low risk accurate 98% of the time, while forecasts that suspects were high risk were accurate 88% of the time. The Hart system was developed in conjunction with the renowned Centre for Evidence-based Policing at the University of Cambridge.
The use of data analytics and AI to help inform police decision making is in line with the Home office’s aspirations outlined in last year’s Modern Crime Prevention Strategy. The Strategy acknowledges that better use of data and technology is one of the key pillars of effective modern crime prevention in the digital age, and outlines the Government’s role in “stripping away barriers to the effective use of data and data analytics, and helping others exploit new and existing technology to prevent crime.”
According to Modern Crime Prevention Strategy data analytics can:
- Help police forces deploy officers to prevent crime in known hotspots (often called ‘predictive policing’)
- Use information shared by local agencies on, for example, arrests, convictions, hospital admissions, and calls on children’s services to identify individuals who are vulnerable to abuse or exploitation
- Spot suspicious patterns of activity that can provide new leads for investigators, such as large payments to multiple bank accounts registered at the same address
- Show which products, services, systems or people are vulnerable to particular types of crime – for example that young women are disproportionately likely to have their smartphone stolen. This means system flaws can be addressed, or crime prevention advice (e.g. on mobile phone security measures) can be targeted more effectively.
SA Company to Use Artificial Intelligence to Predict Crime
Designed to predict and map potential crimes, Solution House Software has announced the launch of their new artificial intelligence (AI) module for Incident Desk. The Incident Desk Predictive Analysis module uses machine learning technology developed by Solution House together with aggregated data from multiple information sources to determine the likelihood of different types of criminal activity in the Incident Desk management area.
“With the module installed, Incident Desk generates 7 and 30-day forecasts as heat maps based on crime types and incident probabilities that managers can use to optimise their finite security resources,” says Janse van Rensburg.
“Crime is notoriously difficult to predict, but given that Incident Desk can access so many different types of data – including weather patterns and forecasts and historical data – the results are based on fairly accurate and proven trending algorithms,” her says.
One of the biggest problems currently plaguing public safety and security are the ‘islands of data’ that are not being shared or centralised, which makes it difficult to data mine and analyse.