Image recognition in child sexual exploitation material—capabilities, ethics and rights
Authors: Phippen, A. and Bond, E.
Image recognition—the use of software to identify images for a specific task—has long been proposed as a solution to addressing some of the more laborious and technical complex tasks such as the triaging of child abuse and exploitation imagery. However, a lack of knowledge and understanding of the capabilities, and limits, of image recognition sometimes results in high levels of expectation that cannot be delivered technically. In tackling key challenges in the policing child abuse material in areas such as victim identification, image categorisation and the prevention of replication and communication of images, image recognition algorithms have been proposed as potential solutions. However, a case study approach highlights concerns around issues such as data bias, lack of training data, and data ethics, which highlight that the complexity of image recognition in this domain might not be as effective as one might wish.