Identify and exploit signatures and observables associated with special nuclear material (SNM) production, storage, and movement.
Challenges of CNEC Thrust Areas
– S&O Thrust Lead
S&O addresses the location of a point source of radiation in an urban environment containing fluctuating background and nuisance sources. S&O is concerned with improving existing and future detector systems by conducting multi-disciplinary research in uncertainty quantification and by analyzing individual sensor systems. In the context of this grand challenge, a signal is defined to be data obtained from a single sensor. Traditionally, a “sensor” in nuclear nonproliferation measures an emission from a material or facility, e.g., ionizing radiation, effluent, or radio frequency. However, we are treating signals in a broader context, to not only include the aforementioned traditional sensors, but also to include data streams such as open-source big data. CNEC faculty students and other researchers are working to identify relevant signatures and observables and conduct basic research on quantifying signal and noise. Since the signal to noise ratio of nuclear proliferation signatures is expected to be small, it is important that clear understandings of signals, noise, and background be developed to address the challenge problem.
Paramount to the solution of any type of signal-to-noise ratio (SNR) problem is developing an accurate estimation of the background. CNEC is working on two major components of feature extraction to address this: 1) nonlinear, feature-based anomaly detection; and 2) knowledge-based background estimation. With these approaches, research focused on developing a sound theoretical framework for S&O emphasizing machine learning approaches is being conducted, including, but not limited to, anomaly detection and its application to high-fidelity feature recovery.