Jan 24, 2022 17:00:00

A project is underway to identify criminals by analyzing 'hands in child sexual abuse images' with AI

Child sexual abuse is a global issue, with the European Parliament passing a bill in 2021 that lets companies check for child pornography in SNS interactions, as well as Apple's announcement of the iPhone. Attempts to prevent sexual exploitation of children by scanning their photos and messages have sparked various debates about privacy and unintended risks . Meanwhile, a research team at Lancaster University in the United Kingdom is working on a project to identify the pedophile who caused the incident from 'hands in pictures of sexual abuse.'

News Item --lend a hand to help crime solving research at lancaster | Student Portal
https://portal.lancaster.ac.uk/portal/news/article/lend-a-hand-to-help-crime-solving-research-at-lancaster

Lancaster project could be used by FBI and Interpol to catch paedophiles --LancsLive
https://www.lancs.live/news/lancashire-news/lancaster-project-could-used-fbi-20969671

Professor Sue Black , a forensic scientist at Lancaster University, has helped arrest many sex offenders by using 'hand-based personal identification techniques.' Black's research shows that the pattern of veins in the hands and wrinkles in the fists is a stronger identifier than DNA, and has helped provide evidence in sex crime cases for nearly 15 years. Black said, 'even the twins are different is the back of the hand,' 'pattern of your veins will be determined at the time of the fetus, individual each to a different anatomical features' and says .

However, it takes a lot of time to examine the hand in the photo taken in the actual sex crime case and compare it with the image of the suspect's hand. So a research team led by Lancaster University lecturer Brian Williams collaborated with Mr. Black to launch a project called 'H-unique'.

You can see what kind of project H-unique is by watching the following movie.

Will you help us to identify child abusers by images of their hands? --YouTube


Child sexual abuse is a global problem ...

According to a 2017 survey, as many as 75,000 web pages contained obscene pictures of children.

In general, the person in the child's sexual image hides his face, but his hands are often not.

Black et al., Who developed a method for identifying individuals from hand images, worked with courts and police to help convict many sex offenders.

However, since the number of specialists who can perform this analysis is small and the time required for one case is enormous, the number of cases that can be handled is limited.

Therefore, a research team at Lancaster University launched a project called 'H-unique' in February 2020 to speed up the process of identifying individuals from hand images.

'H-unique' aims to develop 'AI that identifies individuals from anatomical features such as veins in the hands, wrinkles in the fists, freckles, spots, moles, and scars of burns' by training computers.

Since AI training requires a lot of training data, the 'H-unique' project is

soliciting hand images from people of various nationalities, races, ages and backgrounds.

The project team explains that all data will be anonymized and deleted when the study is complete.

If AI can quickly identify sex offenders, it will be possible to save more children.

As of July 2021, it is said that a total of more than 20,000 images have been collected from about 2,200 people, but when another 2,800 people provide images and the training data is for 5,000 people, a trial Black claims that he can create a sufficiently powerful AI as evidence of.