Home > Seminars > Clare Harshey & Alicia Scott-Automated Cyber Evaluation (ACE) Research Project & Natural Language Processing (NLP) for Reliability Analysis

Clare Harshey & Alicia Scott-Automated Cyber Evaluation (ACE) Research Project & Natural Language Processing (NLP) for Reliability Analysis

Start:

2/7/2019 at 3:30PM

End:

2/7/2019 at 4:45PM

Location:

126 DeBartolo

Host:

College of Engineering close button
headerbottom

Walter Scheirer

Walter Scheirer

VIEW FULL PROFILE Email: walter.scheirer@nd.edu
Phone: 574-631-2436
Website: http://www.nd.edu/~wscheire
Office: 321C Stinson-Remick Hall

Affiliations

College of Engineering Assistant Professor
Primary interests in Computer Vision, Machine Learning, Biometrics,and Digital Humanities. Specific areas of research include Open Set Recognition, Extreme Value Theory Models for Visual Recognition, Biologically-inspired Learning Algorithms, and Stylometry.
Click for more information about Walter
574-631-2436
Add to calendar:
iCal vCal

ACE: The purpose of the Automated Cyber Evaluation (ACE) research project at Naval Surface Warfare Center, Crane Division (NSWC Crane) is to utilize descriptive documentation of United States Marine Corps cyber systems in order to provide the USMC with automated cybersecurity risk evaluations through a natural language processing-informed deep learning model. The ACE model vectorizes raw documentation using Word2Vec and uses the resulting embeddings to train a Convolutional Neural Network, which in turn produces likelihoods for each of five cybersecurity risk labels ranging from Very Low to Very High. The USMC currently employs a time and labor intensive process to manually assess cybersecurity risk for each system. The immediate value of ACE is to provide a “sanity check” for human analysts in order to further standardize understanding of how risk ratings are assigned. Future efforts could alleviate the need for entirely manual assessments and result in significant cost savings to the Department of Defense. This technique could also be transferred to alleviate time and cost constraints in other USMC, Navy and Department of Defense processes.

NLP for Reliability Analysis: Legacy Navy systems collect data in large unstructured formats that are unsuitable to quick programming analyses. In the current process, data takes months of time for engineers to manually sort through in order to categorize, rate urgency, and decide causality. This process is time consuming and labor intensive. The Natural Language Processing (NLP) for Reliability Analysis project at Naval Surface Warfare Center, Crane Division (NSWC Crane) takes the existing unstructured data and applies NLP to analyze the text, pull out the trending topics, and organize by number of occurrences. In the initial research, an analysis that had previously been manually completed by engineers was recreated using the model. What took the engineers 2 months of painstaking work, took the program 11 seconds to process. The program is currently operating with 80% accuracy. The model is using Python programming and making use of open source libraries. As the project expands, the plan is to utilize NLP and machine learning to identify trends, determine causality, and generate notifications when certain threshold levels are encountered.  The application will help move analyses and system performance knowledge closer to real time.

Seminar Speaker:

Clare Harshey & Alicia Scott

Clare Harshey & Alicia Scott

NSWC-Crane Division

Clare Harshey is a software engineer in the Expeditionary Warfare department of NSWC Crane, where she serves as the technical lead for the Automated Cyber Evaluation project through the Crane Artificial Intelligence Development Laboratory. Her background in theoretical linguistics, computer science, and technical editing inform her current interest in research and engineering the fields of natural language processing and human-computer interaction. Clare holds a Bachelor of Arts in Linguistics from the University of Kentucky and a Master of Science in Computational Linguistics from Indiana University, Bloomington.

Alicia Scott is a mathematics and data engineer in the Global Deterrence and Defense Department of NSWC Crane, where she serves as the technical lead for the Natural Language Processing for Reliability Analysis project through the Crane Artificial Intelligence Development Laboratory. Her current research is in advanced data analytics methodology and optimization. Alicia has a Bachelor of Science in Theoretical Mathematics from Indiana University and has 5 years of experience in industry working with statistical analysis and data reconfiguration as well as a variety of data science initiatives. Before joining R&D efforts with the DoD, Alicia advanced to senior leadership for one of the largest company clients and implemented optimization efforts across the department. Now, Alicia’s goal is to use her experience and education to tackle large data science problems for the DoD.