Police badgers and uniforms

Data filter will help police pinpoint the most dangerous online paedophiles

Image credit: Getty Images

Machine learning-type technology unveiled this week in America could also help officers learn how to “better portray” a 10-year-old so as to launch sting operations against would-be sex offenders

American academics are honing a suite of algorithms to help police direct their efforts towards stopping the most dangerous paedophiles.

An automated filtering system was unveiled this week by an online crime expert, Kathryn Seigfried-Spellar, who says the technology can distinguish so-called ‘contact offenders’ – those wishing to harm children in person – from others apparently content to merely imagine physically harming youngsters.

Paedophiles in the latter category are considered to be less dangerous, in the short term at least, than their contact-offending counterparts. While the ‘fantasy-only’ variety will often engage children in sexual conversations online, they mostly appear to be able to draw the line there and are considered comparatively unlikely to commit actual assaults in the near future.

The Chat Analysis Triage Tool (CATT) crunches data harvested from paedophiles’ online conversations and stored on police databases in the USA. The data was provided to a team of American researchers voluntarily by police surveillance teams.

An exploratory study by Seigfried-Spellar and her colleague Professor Ming Ming Chiu examined more than 4,300 messages in 107 online chat sessions involving arrested sex offenders.

The sheer volume of these types of conversations on chat rooms and social networks at any one time means officers are inundated with cases of suspected grooming.

CAAT appears to be a significant step because it could potentially help police to prioritise those cases assessed as most urgently requiring investigation – namely those where a child is at immediate risk of an actual sexual assault.

The tool uses natural language analysis based around use of certain keywords and conversation patterns. While paedophiles intending simply to fuel their fantasies typically engage in relatively brief chats with a succession of different children, would-be contact offenders tend to devote much greater amounts of time – sometimes weeks or even months – to grooming a single individual in order to win their trust and secure a physical meeting.

Aside from the length of online chats, another standout characteristic of the higher risk group of paedophile includes so-called ‘self-disclosure’ behaviour. This is a tactic in which the suspect apparently tries to develop trust, and potentially elicits sympathy, by sharing a private story - normally a negative personal narrative, for example one involving parental abuse.

Seigfried-Spellar explained: If we can identify language differences, then the tool can identify these differences in the chats in order to give a risk assessment and a probability that this person is going to attempt face-to-face contact with the victim.

“That way officers can begin to prioritise which cases they want to put resources toward to investigate more quickly.

It is possible that the machine learning-type technology unveiled by Seigfried-Spellar could in future teach officers how to “better portray” a 10-year-old girl in order to launch ‘sting’ operations, according to a Science Daily notice.

Big data might enable officers to more accurately mimic the syntax and communication styles of children in order to catch those who wish to prey on them.

Four years ago a Dutch charity helped catch 1,000 online sexual predators by using a computer avatar of a child and a fake profile as ‘bait’ on the internet. Terre des Hommes carried out an audacious 10-week sting, posing on video chat rooms as ‘Sweetie’, a 10-year-old Filipina girl.

Some 20,000 men contacted her, with 1,000 found to have offered her money. The names of the men were then passed by the charity to police in the Netherlands, Britain and elsewhere.

According to one estimate by the National Crime Agency, 50,000 people in the UK alone have actively sought out child abuse images online.

This week the Internet Watch Foundation (IWF) revealed that the most up-to-date data showed more of this disturbing material was being found than ever before, with global IWF figures showing it was up by a third. Figures also showed a 37 per cent increase in child sex abuse URLs, the IWF said.

Professor Richard Wortley, director of the Jill Dando Institute of Security and Crime Science, has said creating a hostile environment for sex offenders online is key to protecting children.

Last year he told E&T: “What’s special about the internet is that it gives you that sense of anonymity. This is about looking at how you can take that away. Overseas, I experiment with putting so-called stop pages up, so you’re out there searching for child pornography and suddenly a note comes up from the police saying, ‘We think you’re looking for child pornography. Stop it or we’ll come and get you.’ Sting operations are also quite successful in creating the perception that the internet is not safe.”

Police use of algorithmic tools remains controversial, however, with concerns having been raised in the past about fairness and privacy. But in spite of this, many senior officers feel the technology could have great potential. Earlier this year the UK's most senior counter-terrorism officer said AI-type systems could be used of monitor tens of thousands of suspected jihadis to watch for suspicious new activity.

Recent articles

Info Message

Our sites use cookies to support some functionality, and to collect anonymous user data.

Learn more about IET cookies and how to control them