Search NASA⌕ Search

Engineering topics

Robinson, Sean M.

Publications and source records attributed to Robinson, Sean M..

16 records
All sourcesNASA NTRSDOE OSTI
DOE OSTI2018

Detecting and quantifying materials in containers utilizing an inverse algorithm with adaptive regularization

A method, system and computer program product for automatically detecting and quantifying materials in containers (e.g., luggage, storage containers, equipment or componentry for processing or handling). An X-ray radiograph of a container is generated. A set of materials of interest (e.g., plutonium, steel) is selected to determine if the object comprises or contains materials of interest. An estimate of the areal densities or thicknesses of each of the selected materials of interest is obtained by minimizing an objective function with respect to the areal densities or thicknesses for each of the selected materials of interest. Adaptive regularization is implemented in the objective function to improve optimization results by adding a constraint to the objective function, where the constraint penalizes the objective function for solutions that do not line up with a prior belief about the solution form (e.g., the solution should not be negative or especially noisy).

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2018

Deep Learning for Unsupervised Insider Threat Detection in Structured Cyber Security Data Streams

Analysis of an organization's computer network activity is a key component of early detection and mitigation of insider threat, a growing concern for many organizations. Raw system logs are a prototypical example of streaming data that can quickly scale beyond the cognitive power of a human analyst. As a prospective filter for the human analyst, we present an online unsupervised deep learning approach to detect anomalous network activity from system logs in real time. Our models decompose anomaly scores into the contributions of individual user behavior features for increased interpretability to aid analysts reviewing potential cases of insider threat. Using the CERT Insider Threat Dataset v6.2 and threat detection recall as our performance metric, our novel deep and recurrent neural network models outperform Principal Component Analysis, Support Vector Machine and Isolation Forest based anomaly detection baselines. For our best model, the events labeled as insider threat activity in our dataset had an average anomaly score in the 95.53 percentile, demonstrating our approach's potential to greatly reduce analyst workloads.

Deep Learning, Insider Threat↗
DOE OSTI2017

A single-pixel X-ray imager concept and its application to secure radiographic inspections

Imaging technology is generally considered too invasive for arms control inspections due to the concern that it cannot properly secure sensitive features of the inspected item. However, this same sensitive information, which could include direct information on the form and function of the items under inspection, could be used for robust arms control inspections. The single-pixel X-ray imager (SPXI) is introduced as a method to make such inspections, capturing the salient spatial information of an object in a secure manner while never forming an actual image. The method is built on the theory of compressive sensing and the single pixel optical camera. The performance of the system is quantified here using simulated inspections of simple objects. Measures of the robustness and security of the method are introduced and used to determine how such an inspection would be made which can maintain high robustness and security. In particular, it is found that an inspection with low noise (<1%) and high undersampling (>256×) exhibits high robustness and security.

Arms Control↗
DOE OSTI2017

Effects of Correlated and Uncorrelated Gamma Rays on Neutron Multiplicity Counting

Neutron multiplicity counting relies on time correlation between neutron events to assay the fissile mass, (α,n) to spontaneous fission neutron ratio, and neutron self-multiplication of samples. Gamma-ray sensitive neutron multiplicity counters may misidentify gamma rays as neutrons and therefore miscalculate sample characteristics. Time correlated and uncorrelated gamma-ray-like signals were added into gamma-ray free neutron multiplicity counter data to examine the effects of gamma ray signals being misidentified as neutron signals on assaying sample characteristics. Multiplicity counter measurements with and without gamma-ray-like signals were compared to determine the assay error associated with gamma-ray-like signals at various gamma-ray and neutron rates. Correlated and uncorrelated gamma-ray signals each produced consistent but different measurement errors. Correlated gamma-ray signals most strongly led to fissile mass overestimates, whereas uncorrelated gamma-ray signals most strongly lead to (α,n) neutron overestimates. Gamma-ray sensitive neutron multiplicity counters may be able to account for the effects of gamma-rays on measurements to mitigate measurement uncertainties.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2017

Deep Learning for Unsupervised Insider Threat Detection in Structured Cybersecurity Data Streams

Situational awareness data is a prototypical example of structured heterogeneous streaming data with asynchronous timing that can quickly scale beyond the cognitive power of a human analyst. We present deep learning methods that can train in real time from unlabeled streams and provide probabilistic anomaly scores. We compare our novel deep and recurrent neural network models to three baseline systems using threat detection recall on the CERT Insider Threat Dataset v6.2. In our experiments, our deep neural network and our recurrent neural network systems outperformed all baselines, and we found that using aggregated activity count features yielded the best performance. Our best model produces on average an anomaly score of 95.53 percentile for the day, suggesting it can facilitate a great reduction in analyst workload. Our models are also interpretable, in that they break down the anomaly score into individual features of user behavior, which could further aid analysts in reviewing potential cases of insider threat.

Tuor, Aaron R.↗
DOE OSTI2017

Deep Learning for Unsupervised Insider Threat Detection in Structured Cybersecurity Data Streams

Analysis of an organization’s computer network activity is a key component of early detection and mitigation of insider threat, a growing concern for many organizations. Raw system logs are a prototypical example of streaming data that can quickly scale beyond the cognitive power of a human analyst. As a prospective filter for the human analyst, we present an online unsupervised deep learning approach to detect anomalous network activity from system logs in real time. Our models decompose anomaly scores into the contributions of individual user behavior features for increased interpretability to aid analysts reviewing potential cases of insider threat. Using the CERT Insider Threat Dataset v6.2 and threat detection recall as our performance metric, our novel deep and recurrent neural network models outperform Principal Component Analysis, Support Vector Machine and Isolation Forest based anomaly detection baselines. For our best model, the events labeled as insider threat activity in our dataset had an average anomaly score in the 95.53 percentile, demonstrating our approach’s potential to greatly reduce analyst workloads.

Tuor, Aaron R.↗
DOE OSTI2016

Performance of coincidence-based PSD on LiF/ZnS Detectors for Multiplicity Counting

Abstract: Mass accountancy measurement is a nuclear nonproliferation application which utilizes coincidence and multiplicity counters to verify special nuclear material declarations. With a well-designed and efficient detector system, several relevant parameters of the material can be verified simultaneously. 6LiF/ZnS scintillating sheets may be used for this purpose due to a combination of high efficiency and short die-away times in systems designed with this material, but involve choices of detector geometry and exact material composition (e.g., the addition of Ni-quenching in the material) that must be optimized for the application. Multiplicity counting for verification of declared nuclear fuel mass involves neutron detection in conditions where several neutrons arrive in a short time window, with confounding gamma rays. This paper considers coincidence-based Pulse-Shape Discrimination (PSD) techniques developed to work under conditions of high pileup, and the performance of these algorithms with different detection materials. Simulated and real data from modern LiF/ZnS scintillator systems are evaluated with these techniques and the relationship between the performance under pileup and material characteristics (e.g., neutron peak width and total light collection efficiency) are determined, to allow for an optimal choice of detector and material.

Robinson, Sean M.↗
DOE OSTI2016

Measurement and Comparison of the Light Output of Ni-doped 6LiF/ZnS for Use in Neutron Multiplicity Counting

Abstract–Alternatives to 3He for neutron detection have recently become attractive for safeguards applications. Pacific Northwest National Laboratory is developing a neutron multiplicity counter that is based on 6LiF/ZnS. Some of the properties of this material, such as the scintillation light decay time, can be tuned by doping the material with a small amount of nickel. This doping affects other properties of the material, in particular the time dependence of the scintillation light output. To determine whether the nickel-doped or undoped 6LiF/ZnS material would better suit the neutron multiplicity counter system, a series of experiments compared the use of undoped 6LiF/ZnS and a Ni-doped variant using both small samples and full-scale detectors made using the two materials. Both materials were manufactured by Eljen Technology of Sweetwater, Texas. The Ni-doped variant produced less light than the undoped material and had a shorter decay time. For all choices of detection threshold above the electronic noise, the decrease in light output did not affect the detection efficiency of the system.

Behling, Richard S.↗
DOE OSTI2016

Pulse-Shape Analysis of Neutron-Induced Scintillation Light in Ni-doped 6LiF/ZnS

Abstract–Alternatives to 3He are being investigated for gamma-ray insensitive neutron detection applications, including plutonium assay. One promising material is lithium-6 fluoride with silver activated zinc sulfide 6LiF/ZnS(Ag) in conjunction with a wavelength shifting plastic. Doping the 6LiF/ZnS(Ag) with nickel (Ni) has been proposed as a means of reducing the decay time of neutron signal pulses. This research performed a pulse shape comparison between Ni-doped and non-doped 6LiF/ZnS(Ag) neutron pulses. The Ni-doped 6LiF/ZnS(Ag) had a 32.7% ± 0.3 increase in neutron pulse height and a 32.4% ± 0.3 decrease in neutron pulse time compared to the non-doped 6LiF/ZnS(Ag). Doping 6LiF/ZnS(Ag) with nickel may allow neutron detector operation with improved signal to noise ratios, and reduced pulse pileup affects, increasing the accuracy and range of source activities with which such a detector could operate.

Cowles, Christian C.↗
DOE OSTI2016

Development of a 6LiF/ZnS-based Neutron Multiplicity Counter

Abstract–Neutron multiplicity counters are used in safeguards to provide rapid assay of samples which contain an unknown amount of plutonium in a potentially unknown configuration. A project at PNNL is using regular and nickel-quenched 6LiF/ZnS neutron-scintillator sheets and wavelength shifting plastic for light pipes in place of 3He. A combination of laboratory and modeling work predicts a LiF/ZnS-based system to be able to match or exceed the performance of the best 3He-based systems available. Also, the Ni-quenched material is expected to allow for improved neutron/gamma-ray discrimination at twice the event rate relative to the non-Ni-quenched LiF/ZnS. A new system based on the LiF/ZnS material is under construction and components are being used to optimize the detection efficiency and neutron/gamma-ray discrimination properties. Components of the new system are partially constructed and undergoing performance testing utilizing high-speed digitizers with field programmable gate arrays to perform the neutron/gamma-ray discrimination. The expected performance of the full-scale system is expected to be nearly the same as for 3He-based systems and is due for completion in 2016.

Stave, Sean C.↗
DOE OSTI2016

Improvements in the Method of Radiation Anomaly Detection by Spectral Comparison Ratios

Algorithms for detecting illicit gamma radiation sources must respond to a wide variety of threat signatures and ignore a wider variety of clutter. Previously, we described the Nuisance-rejection Spectral Comparison Ratio Anomaly Detection (N-SCRAD) method for indicating the presence of threat sources during mobile search operations. Our method relies on the formulation of radiation alarm test statistics applied to detector counts in numerous spectral regions of interest and our test statistics are standardized by local estimates of background variability. In this work we present a new automated procedure for selecting test statistics and specifying the regions of interest based on minimizing detectable count rates of prototype threat spectra at a specified false positive rate using simulated annealing. The new procedure can be applied to configuring the detection algorithm when the prototype threats are many and diverse without significantly compromising the detection sensitivity to any one of them. N-SCRAD makes use of estimates of background variability obtained during a search from prior measurements of background. In applications such estimates made purely from past measurements sometimes do not extrapolate well to current conditions. In this work we present a new method for correcting those estimates to current conditions of the total count rate. We demonstrate that the correction lowers detection thresholds for a specified false positive rate, enabling greater sensitivity to targets. The two proposed improvements to N-SCRAD maximize detection power for the expected variety of possible threats while reducing or maintaining the false positive rate.

Anomaly detection, Gamma-ray spectroscopy, Radiati↗
DOE OSTI2015

Combining Radiography and Passive Measurements for Radiological Threat Localization in Cargo

Detecting shielded special nuclear material (SNM) in a cargo container is a difficult problem, since shielding reduces the amount of radiation escaping the container. Radiography provides information that is complementary to that provided by passive gamma-ray detection systems: while not directly sensitive to radiological materials, radiography can reveal highly shielded regions that may mask a passive radiological signal. Combining these measurements has the potential to improve SNM detection, either through improved sensitivity or by providing a solution to the inverse problem to estimate source properties (strength and location). We present a data-fusion method that uses a radiograph to provide an estimate of the radiation-transport environment for gamma rays from potential sources. This approach makes quantitative use of radiographic images without relying on image interpretation, and results in a probabilistic description of likely source locations and strengths. We present results for this method for a modeled test case of a cargo container passing through a plastic-scintillator-based radiation portal monitor and a transmission-radiography system. We find that a radiograph-based inversion scheme allows for localization of a low-noise source placed randomly within the test container to within 40 cm, compared to 70 cm for triangulation alone, while strength estimation accuracy is improved by a factor of six. Improvements are seen in regions of both high and low shielding, but are most pronounced in highly shielded regions. The approach proposed here combines transmission and emission data in a manner that has not been explored in the cargo-screening literature, advancing the ability to accurately describe a hidden source based on currently-available instrumentation.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2015

A Machine Learning Approach for Business Intelligence Analysis using Commercial Shipping Transaction Data

Business intelligence problems are particularly challenging due to the use of large volume and high velocity data in attempts to model and explain complex underlying phenomena. Incremental machine learning based approaches for summarizing trends and identifying anomalous behavior are often desirable in such conditions to assist domain experts in characterizing their data. The overall goal of this research is to develop a machine learning algorithm that enables predictive analysis on streaming data, detects changes and anomalies in the data, and can evolve based on the dynamic behavior of the data. Commercial shipping transaction data for the U.S. is used to develop and test a Naïve Bayes model that classifies several companies into lines of businesses and demonstrates an ability to predict when the behavior of these companies changes by venturing into other lines of businesses.

Bramer, Lisa M.↗
DOE OSTI2015

Neutron pileup algorithms for multiplicity counters

Abstract The shortage of helium-3 (3He) has created a need to identify alternative neutron detection options for a variety of nuclear nonproliferation applications. One application that may be affected by 3He replacement technology is that of mass accountancy for safeguards, which utilizes coincidence and multiplicity counters to verify special nuclear material declarations. The use of neutron scintillation materials, such as LiF-ZnS sheets, as an alternative to 3He proportional tubes in multiplicity counters requires novel techniques for Pulse Shape Discrimination to distinguish between neutrons and gamma rays. These techniques must work under high count rates, as the maximum momentary rate for incoming neutrons from multiplicity events can be quite large. We have created a fast and accurate neutron discrimination algorithm based on time window filtering and signature comparison that can operate quickly on data with high degrees of gamma ray and neutron pileup. This algorithm is evaluated for its capability to separate signals as the pileup rate increases, and the possibility for implementation on fast hardware (e.g., FPGA hardware) for real-time operation is explored.

Robinson, Sean M.↗