State investigative authorities in Germany are constantly and increasingly facing crimes in the area of child abuse or child pornography. The extent of these crimes is increasing significantly, also recognizable by the current press and media coverage: the involvement of large circles of perpetrators, some across national borders, and a frighteningly professional approach. A crucial aspect of this crime is the distribution of child and youth pornography material via data carriers and the Internet. Due to he relatively easy access to appropriate software and hardware produces, huge amounts of data are produced, stored and disseminated.
In the fight against these crimes, the evaluation of the data is a crucial part of the work of the authorities: the filtering out of criminally relevant data with illegal content is the decisive criterion for the prosecution and detention of the perpetrators. Because often the mostly digital image and video material is the most important, if not the only evidence. In addition to managing the large amounts of data, the time factor is crucial: the investigative authorities usually have only a few months to charge a suspect in custody.
In the search for ways to relieve the investigators during the sighting and evaluation, the state of NRW has a study commissioned in 2019, the results of which are now available: the Artificial Intelligence (AI) software developed as a prototype can detect and distinguish with an accuracy of 92% child and adolescent pornography as well as non-criminal adult pornography. Thus, the use of AI means a significant and highly efficient support for the investigators and a great psychological relief. A suitably trained AI can quickly test and sort a large amount of data for the criteria sought with a high degree of accuracy, with the aim of pre-sorting the material for further testing by officials. Time is a decisive factor: the AI of irisnet e. g. creates 30 – 100 frames per second depending on the complexity of the criteria to be tested and depending on the resolution of images.
The AI software of irisnet is now very successfully trained on various so-called checks, this includes Nudity and a distinction between adults and children or e. g. a recognition of age in age groups Dr. Walter Benzing, Founder and Managing Director of irisnet explains: “Our AI now recognizes nudity and other criteria through our longterm high level training with an accuracy of over 95%. This rate far exceeds the hit rate of personnel deployed, as people experience fatigue and therefore mistakes relatively quickly. An AI can go through day and night. Estimating the age is a very special challenge, because in certain phases of life, e. g. in the group of 10-20-year-olds, the exact age is difficult to detect even for people. I am therefore particularly proud of our results in this check criterion, which are around 90%.”
The support of investigative authorities by AI shows a new, very important field of application in which AI technology can prove its clear strengths. However, the crime area of child and adolescent pornography also makes it clear that each AI grows only through the appropriate training material and is continuously improved until its application. irisnet offers to further develop the company’s own AI software together with its customers and to take in account the special customer requirements. We advise our customers for their specific application area and develop targeted trainings until the AI meets our high requirements.
