2013–2014 Dartmouth Computer Science Colloquium Schedule
All colloquia take place on Wednesday at 4:15 in 006 Steele unless otherwise noted.
September 18, 2013:
Wayne Burleson, University of Massachusetts Amherst, Electrical and Computer Engineering
RFID Privacy: From Transportation Payments to Implantable Medical Devices
Although RFID has been widely known for its impact on supply chain and inventory management, two of the most exciting applications from a privacy perspective are in: 1) transportation payment systems and 2) implantable medical devices. This talk presents recent research in both areas, drawing parallels but making important distinctions between the two applications. Both projects involve broad international collaborations due to the large number of technical disciplines involved, as well as varying legal and societal dimensions across different cultures. Transportation payment systems have the ability to divulge user location and hence travel habits. However they also facilitate sophisticated dynamic fare schemes and optimization of the transportation system. Implantable medical devices contain extremely private information about personal health and habits, as well as enabling tracking and other privacy concerns. However, the ability to wirelessly access implanted devices provides enormous health and cost benefits. Both topics raise interesting cross-disciplinary issues in economics, threat models, and ethics as well as more technical aspects of security engineering. This talk will review engineering solutions to both of these domains, including low-power cryptography, physical unclonable functions, and prototyping techniques.
Note: a version of this was presented as the Keynote at RFIDSec 2013 in Graz, Austria
Wayne Burleson has been a Professor of Electrical and Computer Engineering at the University of Massachusetts Amherst since 1990. He is also currently a Senior Fellow at AMD Research in Boston. He has degrees from MIT and the University of Colorado. He has worked as a custom chip designer and consultant in the semiconductor industry with VLSI Technology, DEC, Compaq/HP, Intel, Rambus and AMD, as well as several start-ups. Wayne was a visiting professor at ENST Paris in 1996/97, at LIRM Montpellier in 2003 and at EPFL Switzerland in 2010/11. His research is in the general area of VLSI, including circuits and CAD for low-power, interconnects, clocking, reliability, thermal effects, process variation and noise mitigation. He also conducts research in hardware security, reconfigurable computing, content-adaptive signal processing, RFID and multimedia instructional technologies. He teaches courses in VLSI Design, Embedded Systems and Security Engineering. Wayne has published over 180 refereed publications in these areas and is a Fellow of the IEEE for contributions in integrated circuit design and signal processing.
October 16, 2013:
Michael Cohen, Microsoft Research
Still, In Motion
This talk will focus on the creation of media that lie somewhere between objective and subjective capture of photographic reality (but closer to objective) and somewhere between still and video (but closer to still). Arriving at this point will require some retrospection on a path from Art through Computer Graphics and arriving at Computational Photography. I will touch on research from the past three decades that combine algorithmic methods from computer graphics and computer vision leading to recent efforts to bring this magic to the devices we carry with us. I'll end with a glimpse into a current project that seeks to combine a 20 Gigapixel panorama of Seattle with artists, dancers, and acrobats.
Michael Cohen is Principal Researcher at Microsoft Research and Affiliate Professor at the University of Washington. Michael has a BA in Studio Art from Beloit College, a BS in Civil Engineering from Rutgers University, an MS in Architectural Science and Computer Graphics from Cornell University and a PhD in Computer Science from the University of Utah. Winner of the SIGGRAPH Computer Graphics Achievement Award (1998) and an ACM Fellow (2007), Michael is a leader in digital arts, computer graphics, and computational photography research. Fourteen of Prof. Cohen's papers have over 300 citations (one with 2200 citations). He has 75 patents and has served as Program, Committee, and Awards Chair for SIGGRAPH and Eurographics. For a complete list of papers and publications see http://scholar.google.com/scholar?q=michael+f+cohen&hl=en&as_sdt=0,48. Michael has also been a graduate advisor, and he mentored many computer graphics faculty and researchers, including Julie Dorsey, Ankit Gupta, Mira Dontcheva, Steve Gortler, Charles Rose, Eric Chen, Paul Isaacs, and Pravin Bhat.
October 23, 2013:
Deepak Ganesan, University of Massachusetts Amherst
Extending our understanding of human behavior through continuous sensing
Our ability to continuously monitor activities, health, and lifestyles of individuals using sensors has reached unprecedented levels --- on-body sensors enable continuous sensing of our physiological signals, smartphones have a plethora of sensors to monitor activity and location, and a growing number of sensors embedded in the physical world enable monitoring of our living spaces. Such ubiquitous sensing promises to revolutionize our understanding of the social, environmental, and behavioral determinants of a wide range of human activities and health conditions.
Despite its promise, there are fundamental challenges in designing such systems in terms of data processing, sensing, and power. How can we make reliable inferences despite the noisy, uncertain nature of natural environments? How can we expand our understanding of human behavior through more sensors that fully capture our actions, attention, and environmental cues? How can we cope with the burden of having to re-charge a growing ecosystem of wearable sensors?
My talk discusses our ongoing work to address these challenges. From a data perspective, I will talk about leveraging machine learning techniques to detect use of addictive drugs with wearable ECG sensors, and methods to fuse information across diverse continuous sensor sources. From a sensing perspective, I will talk about the design of computational eyeglasses, a wearable sensor that continuously tracks eye and visual context. From a power perspective, I will discuss our work on RF-powered sensor devices that can sense, process and communicate at orders of magnitude less power than a typical battery-powered sensor.
Deepak Ganesan is currently Associate Professor in the Department of Computer Science at UMASS Amherst. He received his Ph.D. in Computer Science from UCLA in 2004 and his bachelors in Computer Science from IIT, Madras in 1998. He received the NSF CAREER Award in 2006, the IBM Faculty Award in 2008, and a UMass Lilly Teaching Fellowship in 2009. His publications have received awards at various conferences, most recently, a Best Paper Award at ACM CHI 2013, and an Honorable Mention for Best Paper Award at ACM Ubicomp 2013. He was a Program co-chair for ACM SenSys 2010 and IEEE SECON 2013.
October 30, 2013:
Aleksandar Kuzmanovic, Northwestern University
Mosaic: Quantifying Privacy Leakage in Mobile Networks
In this talk, accessible to everyone, I will present a measurement study on privacy challenges in mobile networks. In particular, with the proliferation of online social networking (OSN) and mobile devices, preserving user privacy has become a great challenge. While prior studies have directly focused on OSN services, we call attention to the privacy leakage in mobile network data. This concern is motivated by two factors. First, the prevalence of OSN usage leaves identifiable digital footprints that can be traced back to users in the real-world. Second, the association between users and their mobile devices makes it easier to associate traffic to its owners. These pose a serious threat to user privacy as they enable an adversary to attribute significant portions of data traffic including the ones with no identity leaks to network users’ true identities. To demonstrate its feasibility, we develop the Tessellation methodology. By applying Tessellation on traffic from a cellular service provider (CSP), we show that up to 50% of the traffic can be attributed to the names of users. In addition to revealing the user identity, the reconstructed profile, dubbed as “mosaic,” associates personal information such as political views, browsing habits, and favorite apps to the users.
Aleksandar Kuzmanovic is an Associate Professor in the Department of Electrical Engineering and Computer Science at Northwestern University. His research interests are in the area of computer networking with emphasis on design, measurements, analysis, denial-of-service resiliency, and prototype implementation of protocols and algorithms for the Internet. He is most well-known for his pioneering work on low-rate denial-of-service attacks, TCP congestion control, and data sharing methods for Internet-scale systems. He has published more than 40 research papers in the most prestigious networking journals and conferences. He received the NSF CAREER Award. He is a member of the steering committee of the Measurement Lab, an open platform for monitoring net neutrality. He is a member of the advisory board of Narus, Inc and a co-founder of several startups.
November 6, 2013:
Gwen Spencer, Dartmouth
Influence Beyond Exposure: Tackling an Economic Variant of Seeding Viral Spread
The problem of seeding "contagion" in social networks has attracted substantial attention for its connection to viral marketing. Simple information-spread models yield nice mathematical properties that allow theoretical algorithmic traction. This work often points to some form of "exposure" as the best paradigm for designing seed sets. While exposure-based seeding may spread awareness effectively, being aware of a behavior often doesn't result in a decision to adopt it. When environmental economists describe decisions to engage in green behaviors, when sociologists model norm-spread, and when game theorists consider choices to adopt cooperative strategies in repeated-game-play, a common more-complex spread mechanism emerges. What can we learn about how to virally market biking to work, adopting health-related behaviors, and cooperating at an equilibrium that is mutually beneficial?
I'll mention convergence results for this spread mechanism, (daunting) hardness results for the seeding question, and (heartening) computational results from heuristics derived by truncating an exact (inefficient) Integer Program. Compared to exposure-based seeding, the advantage of a seeding paradigm that establishes "critical mass locally" appears largest when the network is highly clustered (as social networks often are).
Gwen Spencer is a Neukom Postdoctoral Fellow, joint in Computer Science and Environmental Studies. Her PhD focused on approximation algorithms for stochastic graph-fragmenting problems motivated by planning spatial placement of controlled burns to limit wildfire damage (at Cornell, with David Shmoys). Gwen is generally interested in optimization problems in networks which consider probabilistic inputs, probabilistic effects of actions, and incentives.
November 13, 2013:
Erik Learned-Miller, University of Massachusetts Amherst
Face Recognition by Computer: Does it work?
In this talk, I will review the broad area of face recognition by computer, including problems such as face detection, face verification, and face identification. The perceptions about how well face recognition works vary widely. I'll try to provide some background for understanding what's really going on, both in research and in industry, and what the state-of-the-art is. I'll also discuss comparisons with human performance in face recognition.
I will review a variety of recent techniques for solving face recognition problems, some developed at UMass Amherst and some developed elsewhere, and give a sense of where current research is headed. Particularly interesting are the errors produced by current face recognition systems. We will conclude that the question of whether face recognition "works" is not meaningful without specifying the context in which it is done.
Erik G. Learned-Miller is an Associate Professor of Computer Science at the University of Massachusetts, Amherst, where he joined the faculty in 2004. He spent two years as a post-doctoral researcher at the University of California, Berkeley, in the Computer Science Division. Learned-Miller received a B.A. in Psychology from Yale University in 1988. In 1989, he co-founded CORITechs, Inc., where he co-developed the second FDA cleared system for image-guided neurosurgery. He worked for Nomos Corporation, Pittsburgh, PA, for two years as the manager of neurosurgical product engineering. He obtained M.S. (1997) and Ph. D. (2002) degrees in Electrical Engineering and Computer Science from the Massachusetts Institute of Technology. In 2006, he received an NSF CAREER award for his work in computer vision and machine learning. He is a Program Chair for the 2015 IEEE Conference on Computer Vision and Pattern Recognition.
November 20, 2013:
Jayant Madhavan, Google
Big Data for Data Enthusiasts
Data enthusiasts, such as data journalists, social scientists and data activists, through their data-driven advocacy play an important role in the big data movement. While they have interesting datasets, they often lack the technical expertise and resources to deal effectively with them. However, much of the current buzz around big data has focused on computational challenges, and ignores the challenges faced by data enthusiasts. In this talk, I will describe how Google Fusion Tables (GFT) is beginning to address their specific data management needs. In particular, I will address two challenges. First, GFT enables users to create and publish interactive visualizations over large datasets, a feature that is key to data-driven storytelling. Interactive maps created using GFT are frequently featured in prominent news stories that generate extremely high volumes of traffic. I will describe the challenges and solutions in supporting such maps over large complex geo-spatial datasets. Second, GFT enables users to find interesting public datasets and to piece together new datasets by combining related ones, features that are key to data advocacy. I will describe the challenges and solutions in finding the right dataset and how such data search can ultimately have a broader impact on web search.
Jayant Madhavan is a member of the Structured Data Research group at Google Inc. His research interests include data integration, web search and data visualizations. Until recently, he led the engineering team for Google Fusion Tables, a cloud data management solution. He was the Chief Architect at Transformic Inc., a portal that built search engines for the Deep Web, which was acquired by Google in 2005. He is a recipient of the Ten Year Best Paper Award at VLDB 2011. He received a Ph.D. from the University of Washington in 2005 and a B.Tech. from IIT Bombay in 1999.