Recent News

	(11/09/09) Prof. Sean Smith and his group will be playing a major in an initiative to secure the power grid. Read more here.
	(11/08/09) Prof. Hany Farid finds that iconic Oswald photo was not faked. Read more here.
	(11/08/09) Prof. Tanzeem Choudhury has been named a TEDIndia Fellow for 2009. Read more here.

Recent Technical Reports

• September 2009: Katana: A Hot Patching Framework for ELF Executables
• October 2010: Detecting Photographic Composites of Famous People
• September 2009: Activity-Aware Electrocardiogram-based Passive Ongoing Biometric Verification
• August 2009: Semantic and Visual Encoding of Diagrams
• July 2009: Distributed Monitoring of Conditional Entropy for Network Anomaly Detection

Featured Research

Learning Human Behavior from Multi-modal Sensor Data

Prof. Tanzeem Choudhury

Systems that can capture and recognize human behaviors in unconstrained real world settings have the potential to dramatically impact research in smart environments, novel user interfaces, surveillance, and health care. For these systems to be practical in the "real world" (i.e., outside of research labs), they need to detect a variety of activities that are performed routinely in many different manners, under many different environmental conditions, and across many different individuals. We are developing machine learning techniques for feature selection, semi-supervised, and unsupervised learning in structured models, which can handle the complexities of real world data, are computationally tractable, and minimize the effort required from humans to train activity recognition systems