Problems and mitigation strategies for developing and validating statistical cyber defenses

The development and validation of advanced cyber security technology frequently relies on data capturing normal and suspicious activities at various system layers. However, getting access to meaningful data continues to be a major hurdle for innovation in statistical cyber defense research. This paper describes the data challenges encountered during development of the machine learning approach called Behavior-Based Access Control (BBAC), together with mitigation strategies that were instrumental in allowing R&D to proceed. The paper also discusses results from applying a spiral-based agile development process focused on continuous experimental validation of the resulting prototype capabilities.