PRODIGAL (computer system)

PRODIGAL (proactive discovery of insider threats using graph analysis and learning) is a computer system for predicting anomalous behavior among humans, by data mining network traffic such as emails, text messages and server log entries.[1] It is part of DARPA's Anomaly Detection at Multiple Scales (ADAMS) project.[2] The initial schedule is for two years and the budget $9 million.[3]

Proactive discovery of insider threats using graph analysis and learning
Establishment2011
SponsorDARPA
Value$9 million
GoalRapidly data mine large sets to discover anomalies

It uses graph theory, machine learning, statistical anomaly detection, and high-performance computing to scan larger sets of data more quickly than in past systems. The amount of data analyzed is in the range of terabytes per day.[3] The targets of the analysis are employees within the government or defense contracting organizations; specific examples of behavior the system is intended to detect include the actions of Nidal Malik Hasan and WikiLeaks source Chelsea Manning.[1] Commercial applications may include finance.[1] The results of the analysis, the five most serious threats per day, go to agents, analysts, and operators working in counterintelligence.[1][3][4]

Primary participants

See also

References

  1. "Video Interview: DARPA's ADAMS Project Taps Big Data to Find the Breaking Bad". Inside HPC. November 29, 2011. Retrieved December 5, 2011.
  2. Brandon, John (December 3, 2011). "Could the U.S. Government Start Reading Your Emails?". Fox News. Retrieved December 6, 2011.
  3. "Georgia Tech Helps to Develop System That Will Detect Insider Threats from Massive Data Sets". Georgia Institute of Technology. November 10, 2011. Retrieved December 6, 2011.
  4. Storm, Darlene (December 6, 2011). "Sifting through petabytes: PRODIGAL monitoring for lone wolf insider threats". Computer World. Archived from the original on January 12, 2012. Retrieved December 6, 2011.


This article is issued from Wikipedia. The text is licensed under Creative Commons - Attribution - Sharealike. Additional terms may apply for the media files.