Dynamic information source selection for intrusion detection systems

Our work presents a mechanism designed for the selection of the optimal information provider in a multi-agent, heterogeneous and unsupervised monitoring system. The selfadaptation mechanism is based on the insertion of a small set of prepared challenges that are processed together with the real events observed by the system. The evaluation of the system response to these challenges is used to select the optimal information source. Our algorithm uses the concept of trust to identify the best source and to optimize the number of challenges inserted into the system. The mechanism is designed for intrusion/fraud detection systems, which are frequently deployed as part of online transaction processing (banking, telecommunications or process monitoring systems). Our approach features unsupervised adjustment of its configuration and dynamic adaptation to the changing environment, which are both vital for these domains.