Safeguards in a Big Data World

Size: px

Start display at page:

Download "Safeguards in a Big Data World"

Alexina Berry
5 years ago
Views:

1 Safeguards in a Big Data World August 2017 Dr. Karen Miller INMM Novel Technologies Workshop Albuquerque, New Mexico Operated by Los Alamos National Security, LLC for the U.S. Department of Energy's NNSA UNCLASSIFIED // LA-UR

Bottom Line The IAEA is fundamentally a data-driven agency Increasing mandate + static budget = need for productivity gains Data science and automation offer potential gains in more

2 Bottom Line The IAEA is fundamentally a data-driven agency Increasing mandate + static budget = need for productivity gains Data science and automation offer potential gains in more efficient safeguards implementation Bringing modern data analytic tools to the IAEA will require collaboration between data scientists and subject-matter experts in safeguards 10/3/2017 2

3 Significant Digits 10/3/2017 3

4 A fresh look at old questions vujá dé Can the IAEA better optimize its limited resources? Can the IAEA do more with the data it s already collecting? Can new approaches enable more efficient safeguards implementation? 10/3/2017 4

Disruptive Innovative Data science for safeguards has more potential to revolutionize the field than any other technology Automation will be key to closing the gap between the amount of data the IAEA

5 Disruptive Innovative Data science for safeguards has more potential to revolutionize the field than any other technology Automation will be key to closing the gap between the amount of data the IAEA can collect and the amount of data it can use IAEA is taking proactive steps to modernize its IT infrastructure How well do we understand the data landscape? Can we identify pain points and specific elements of IAEA workflows that can be streamlined? The Innovation S-curve 10/3/2017 5

Traditional Safeguards Nondestructive assay systems Cross-correlation between neutron and gamma-ray measurements Machine learning for spent fuel measurements Advanced algorithms for list mode data

6 Traditional Safeguards Nondestructive assay systems Cross-correlation between neutron and gamma-ray measurements Machine learning for spent fuel measurements Advanced algorithms for list mode data analysis Computer vision and image analysis Spent fuel Cerenkov imaging systems Verifying authenticity of tags and seals Automated data labeling for surveillance videos Remote and unattended monitoring systems Leverage advances in commercial IoT technologies Advanced statistics Bayesian methods for error analysis Advances in near-real-time accounting for process monitoring 10/3/2017 6

7 Open Source Analysis Safeguards implications of the small satellite revolution? Lower resolution but global daily coverage Anomalous change detection for automated cueing of inspections Text analytics Data triage and reduction Automated categorization and geo-tagging of technical documents Disparate data integration Single data streams often have limitations The whole is greater than the sum of its parts Recognizing synthesized activities across disparate data streams 10/3/2017 7

Human-Machine Interfaces Data visualization provides the mechanics to compress a significant amount of information Enables knowledge discovery Critical for making sense of complex

8 Human-Machine Interfaces Data visualization provides the mechanics to compress a significant amount of information Enables knowledge discovery Critical for making sense of complex data Augmented reality Retrieve important information on headsup display during inspection Overlay of sensor data Increased emphasis on mobile computing and user experience 10/3/2017 8

Data Authentication Data authentication at the sensor Elimination of the local PC for remote monitoring systems Data security and transfer directly at the sensor level Does blockchain have a role to

9 Data Authentication Data authentication at the sensor Elimination of the local PC for remote monitoring systems Data security and transfer directly at the sensor level Does blockchain have a role to play in safeguards? Distributed leger technology Implications far beyond cryptocurrency Provides tamper indication and authentication of transactions Immutable records for supply chain track and trace 10/3/2017 9

10 Conclusions The first 60 years of safeguards created an incredible springboard from which to launch the safeguards technologies of the future The world is at an inflection point where our ability to collect data now far outpaces our ability to make use of it The convergence of existing safeguards expertise with modern data science has the potential to generate revolutionary capability advances for the IAEA The most exciting work is still to come 10/3/

CVT Workshop October 31 November 1, 2018

CVT Workshop October 31 November 1, 2018 Anomaly Detection in the Monitoring of Nuclear Facilities Elizabeth Hou, Karen Miller, Alfred Hero University of Michigan, LANL, University of Michigan 11/01/2018