Computer scientists have devised a new technique for analyzing image and video material for child pornography material. By simply studying noise it is possible to identify child abuse. International investigative organizations have already shown interest.
This is how the new detection method works
Photos and videos are never perfect: they always contain a speck or other form of noise somewhere. This is not always visible to the naked eye, but it is indeed present. This is due to the production process of the image sensor.
These small imperfections on the image sensor are unique. If you know the noise pattern of a sensor, you can link photos and videos. You can compare it to ballistic research. Each firearm leaves a distinctive pattern when a bullet is fired. Using that pattern, forensic experts can link a bullet to a crime scene and link a weapon.
Much interest in the Dutch investigation model
Guru Bennabhaktula, a PhD student at the University of Groningen and the Spanish University of León, devised and developed a machine learning model for his PhD research to extract and analyze the noise from image and video material. His model can be used as a new investigative tool for investigative and enforcement authorities.
For example, if officers find a camera on a suspect, detectives can investigate whether the device was used to witness child pornography. The image and video material can be linked to photos and videos that have been shared and distributed on the internet or the dark web.
Bennabhaktula says Interpol has already expressed interest. “We also