Towards Automatic Threat Recognition

The current transformation processes aim at robustly networked forces in order to enable them to execute network-centric operations [1]. Obviously, this will provide the forces ’ headquarters with a huge amount of data and information that has to be processed to deduce an appropriate picture of the respective battlespace and evolving threats in a timely and most reliable manner. Some of the data will not be relevant at all, but some other may indicate upcoming threats. This is true especially with respect to reconnaissance reports which will not only include information from own reconnaissance assets but also from civilians and other open sources. Thus, one of the most challenging aspects of data and information processing in military affairs is to find the situation relevant information within the huge amount of irrelevant one. The paper at hand describes a high level information fusion system under development. This system automatically processes the incoming stream of unstructured information in order to support the human operator in intelligence processing. Ideally, analysing the input information based on the so far perceived situation and available background knowledge, the system recognises that part of information which might indicate new threats (as well as targets) or confirm existing hypotheses. The relevant and pre-processed information then will be presented to the operator for further (interactive) investigation. Our system processes the incoming messages in two steps according to the established intelligence processing procedures. During the first one, the reported events are categorised (or classified), and cross-referenced. In the second one, the resulting information is analysed and combined and integrated into patterns in the course of the production of further intelligence in order to separate the presumable relevant information from the less significant one. For both steps, knowledge about the domain and the information context has to be accessed and exploited. Therefore, the system resorts to an ontology. To point out the system and its functionalities, the paper will describe this ontology as well as the processing steps in general and by example