Hand ID project aims to catch child abusers in the act
Image credit: motortion | Dreamstime
Scientists behind a hand-identification research programme are launching a new app, with hopes that a large collection of images submitted to the platform will help investigators identify child abuse perpetrators from footage shared online.
The researchers, from Lancaster University and the University of Dundee, are calling on thousands of members of the public to submit pictures of their hands and wrists into the app to help train a computer system that will extract anatomical features of hands from anonymous photographs. These images will allow the team to design algorithms that will help police to link suspects to crimes just from pictures of their hands.
“We can use this knowledge to develop sophisticated computer algorithms and new forensic tools that will help law enforcement apprehend those who harm the most vulnerable in our society,” said forensic anthropologist Professor Dame Sue Black.
“But we can’t do this without the help of thousands of volunteers,” she disclosed. “This vitally important work depends on our being able to analyse a large number of hands to see what differences there are.”
Led by Professor Black, the H-Unique programme aims to discover whether our hands are truly unique by looking at anatomical differences caused by development, genetics, ageing, the environment and even accidents.
“We know that features such as vein patterns, skin creases, freckles, moles, and scars are different between our right and left hands, and even differentiate between identical twins,” said Black. “We are looking to deliver a step-change in the science so we can analyse and understand all factors that make a hand unique.”
A prime motivation for the study is to help investigators identify child abuse perpetrators from footage and images shared online, where backs of hands are often one of the only visible features of the abuser.
According to researchers, the approach could become an invaluable new tool for informing criminal courts and thereby give juries a greater degree of certainty in their deliberations on whether to convict or acquit those accused of some of the most heinous crimes.
The project combines ‘hard biometrics’, such as fingerprints with ‘soft biometrics’ to obtain a comprehensive assessment of the person’s true identity.
“The tools we will develop will reliably and robustly inform decisions in criminal courts,” said Dr Bryan Williams, lecturer in biometrics and human identification at Lancaster University. “They could also be used to assist law enforcement agencies to rapidly and autonomously analyse hours of footage and thousands of offensive images.”
Training the computers requires a large number of photographs of hands and the team are seeking help from anyone over the age of 18 – from all ethnicities, nationalities and backgrounds. They also said more than 5,000 ‘citizen scientists’ are needed to contribute – so that there is enough data to prove beyond a reasonable doubt whether our hands are unique.
A web-based app for anyone aged 18 and over to contribute their images to the project is available to use on smartphones at h-unique.lancaster.ac.uk.
The team said the app, which only takes around 10 minutes to use, provides clear instructions on how to take images from angles they require. Such images are then sent anonymously to the research team and used as part of a database for developing hand comparison algorithms.
Williams explained: “The app offers simple step-by-step instructions explaining the kinds of images and different angles we need of each hand.
“These anonymous images will be marked up by experts for potentially distinct features based on existing anatomical knowledge. These will be compared to check that no two hands are exactly the same.
“We will also develop computer models based on mathematics and computer science that we will train to reliably and repeatedly extract anatomical information regardless of conditions and even when hands are not in ideal positions.”
The team said the images collected will not be shared with any external agencies and will be destroyed at the end of the five-year research project, funded through a €2.5m (£2.1m) grant from the European Research Council.
Sign up to the E&T News e-mail to get great stories like this delivered to your inbox every day.