of PREVISION is to support law enforcement agencies in the collection and analysis of large, heterogeneous data sets using modern artificial intelligence methods. The resulting improvement in the prevention and investigation of crimes in the area of organized crime contributes to increasing public security.
Ensuring public security presents law enforcement agencies with the challenge of collecting and analyzing increasingly large amounts of data. Specifically, this involves text and message formats such as witness statements, emails, or social network posts but also images, videos, telecommunications and traffic data, and several others. In particular, data that is generated by machines, such as videos from surveillance cameras, accumulates in such large masses that a comprehensive evaluation is only possible with machine support. PREVISION aims to provide the necessary technical tools for this purpose in close cooperation with law enforcement agencies.
In PREVISION, ten institutions with police and legal backgrounds, such as law enforcement agencies focusing on counter-terrorism or personal protection, police authorities, police schools, ethics experts and sociologists, are working together with eighteen project partners with technical expertise, including universities, research centers and software companies, in an interdisciplinary consortium from thirteen European countries.
Based on five selected deployment scenarios from the application areas of counterterrorism, white-collar crime, and art smuggling, the project will first identify the need for solutions, specify technical requirements, concretize ethical and legal guidelines, and discuss the impact of crime and crime-fighting methods on society.
Based on this, established algorithms of text, image and audio processing will be tailored to the respective use case as well as existing methods of artificial intelligence will be further developed. A core objective of the research project is the semantic merging of manually and machine-derived facts in an ontology-based knowledge model. By using methods of data classification, machine reasoning and data clustering, information hidden in the data stock can be revealed and temporal trends for the development of threads can be derived.
A special challenge is the conception of the functionality described above in a modular, standardizable and thus future-proof architecture with open interfaces, which favors the emergence of a new market for surveillance technologies.