Rutgers-led center to tease out terrorism clues, detect disease outbreaks
A new Rutgers-led center funded by a $3 million, three-year Department of Homeland Security (DHS) grant officially opened for business last week. Amid the celebration, however, was a healthy dose of introspection about how to protect peoples’ privacy as researchers work to ferret out terrorism clues or spot early warnings of disease outbreaks.
The Center for Dynamic Data Analysis (DyDAn), announced last summer, is one of six national DHS university "centers of excellence." It will develop computing technologies that find patterns and relationships in publicly accessible writings, such as news stories and web logs, to quickly identify emerging indicators of possible terrorist activity. The technologies also will rate the consistency and reliability of the information sources. Such information could give officials more lead time to investigate and potentially thwart terrorist plans.
The center also will serve as an information management resource for other homeland security concerns, such as epidemic disease outbreaks, natural disaster response, cargo container inspections, and accidental import of invasive plant and animal species.
While DyDAn’s emphasis is on examining public data, center director and math professor Fred Roberts told a group of faculty, students, and corporate and government partners attending opening ceremonies about the need to protect privacy while they work to uncover terrorists’ rapidly changing tactics or shifting occurrences of diseases. To help better understand the privacy issue and other technical challenges, Roberts invited three speakers to share their insights.
Keynoting the meeting was Stephen Fienberg, professor of statistics and social science at Carnegie Mellon University, and widely considered one of the field’s most influential statisticians.
Record linkage, or the ability of one database to share and connect information with other databases, is especially prone to privacy abuse, Fienberg said. “Masses of transaction data are being collected daily, from E-ZPass to online banking,” he said. “Who controls what and who has access to what?” Combining this information enables companies or governments to know more about people than they need to know. And worse, it can easily result in incorrect inferences. Fienberg discussed techniques for guarding against making inappropriate linkages, from data encryption to techniques that perform data matches behind virtual barriers that selectively reveal results.
Also addressing the group were Joseph Kielman, director of research futures for the Department of Homeland Security, and James McGraw, deputy director of the Institute for Scientific Computing Research at Lawrence Livermore National Laboratory (LLNL). Kielman said he views data as a perishable commodity, which the government needs to analyze and act on as it pours in, rather than collecting it first and running analyses later. McGraw noted the need for computer systems to do a preliminary examination and prioritization of information presented to analysts, who simply can’t handle the volume of raw data that pours in. The challenge to DyDAn and other university groups working with LLNL on these issues is to develop ways to match patterns in large, rapidly changing collections of data and determine how strongly those patterns match.
DyDAn is based at Rutgers’ Center for Discrete Mathematics and Theoretical Computer Science (DIMACS), also headed by Roberts. It is a collaboration of five university and two industry partners. DyDAn also coordinates work by three other university-led consortia doing data analysis research. All work with the Department of Homeland Security through Lawrence Livermore’s Institute for Discrete Sciences.