NetSE: Medium: A Data Mining Approach to Diagnostic Debugging in Sensor Networks

点击翻译按钮获取中文摘要

Abstract (limited to 250 words): This award is funded under the American Recovery and Reinvestment Act of 2009 (Public Law 111-5). Our nation's infrastructure relies increasingly on networks that connect growing amounts of data and systems, including those that interact directly with the physical world. Increased connectivity leads to a higher degree of vulnerability to attacks, malfunctions, and failures that can cascade more rapidly along network links. The project develops techniques to improve the reliability of emerging networked infrastructure, where computation, communication, and sensing are intimately intertwined. Use of data mining techniques is investigated to determine and eliminate scenarios involving cascaded failures and propagation of performance problems. The complexity of emerging networked and pervasive computing systems increases maintenance cost, challenges classical design approaches, and makes traditional diagnostics and debugging tools less effective at catching problems. To reverse these trends, this project develops tools that are specifically suited to address three fundamental challenges of complex distributed systems; namely, non-reproducible stochastic behavior, high interactive complexity, and physical resource constraints. Other than improving reliability, this research is integrated with education curricula at the University of Illinois, offering real-world challenges to intellectually stimulate both graduate and undergraduate students, while seeking avenues to encourage cultural diversity and promote women and minority involvement in engineering. Laboratory modules allow students to experiment with and diagnose real-world design problems and cascading interaction anomalies in a hands-on fashion. The project will result in improved versions of a data mining textbook by the Co-PI, which is currently considered the standard reference in the field.