- talks - lists.cs.princeton.edu

Talk starting now in small auditorium
by Jennifer Rexford 09 Dec '13

09 Dec '13

Sent from my iPad

1 0

Colloquium Speaker Yuanyuan Zhou
by Nicole E. Wagenblast 09 Dec '13

09 Dec '13

Do not blame users for misconfigurations Yuanyuan Zhou , University of California, San Diego Monday, December 9, 4:30pm Computer Science 105 Today's data centers usually employ high redundancy to tolerate hardware and software errors. But unfortunately another type of errors, namely configuration errors (i.e., misconfigurations) can still take down the whole data centers. It has contributed to more than 30% high severity issues and caused some of the recent major downtime in cloud service companies including Amazon, Microsoft Azure, Google, Facebook, etc. Unfortunately, many software developers and system designers put most of blames on users for configuration errors, and do not pay enough attention to handling misconfiguration in a more active way. Comparing to software bugs, configuration issues have much less tooling support for error detections, issue tracking, tolerance testing, as well as design reviews. In this talk, I will present our recent work on characterizing misconfigurations in commercial and open source systems (as well as a major commercial cloud service provider), and also some of our solutions in addressing this configuration problem. We approach it from the perspective that configurations are a part of user interface, and thereby need to consider from user perspectives. More specifically, as a practical first step, we need to avoid error-prone requirements and also react gracefully to user mistakes. Our solutions have been used two commercial companies and have influenced some popular open source systems such as "Squid" redesigning their configuration. In this talk, I will also some of the negative interactions (and "challenges") with some open source developers. Yuanyuan Zhou is currently a Qualcomm Chair Professor at UC-San Diego. Before UCSD, she was a tenured associate professor at University of Illinois at Urbana Champaign. Her research interests span the areas of operating systems, software engineering, system reliability and maintainability. She has 3 papers selected into the IEEE Micro Special Issue on Top Picks from Architecture Conferences, best paper at SOSP'05, and 2011 ACM SIGSOFT FSE (Foundation of Software Engineering) Distinguished Paper. She has co-founded two startups. Her recent startup, PatternInsight, has successfully deployed software quality assurance tools in many companies including Cisco, Juniper, Qualcomm, Motorola, Intel, EMC, Lucent, Tellabs, etc. Recently Pattern Insight has sold its Log Insight business to VmWare. In addition, Intel has licensed some of her solutions on detecting concurrency bugs. She has graduated 14 Ph.D. students so far, many of whom have joined top industrial companies and academia such as University of Wisconsin at Madison, University of Toronto, University of Waterloo, etc as tenure-track faculty.

1 1

Fwd: Adi's talk Monday
by Moses S. Charikar 05 Dec '13

05 Dec '13

Adi Shamir will be speaking at IAS this coming Monday, Dec 9. Details below: Cheers, Moses How Cryptosystems Are REALLY Broken Computer Science/Discrete Mathematics Seminar I Topic: How Cryptosystems Are REALLY Broken Speaker: Adi Shamir Affiliation: Weizmann Institute Date: Monday, December 9 Time/Room: 11:15am - 12:15pm/S-101 Most of the cryptosystems we currently use are highly secure, and cannot be broken with reasonable complexity by mathematical cryptanalysis. However, over the last fifteen years researchers have developed many types of physical attacks on their implementations which can easily bypass their mathematical security. In this talk I will survey some of the latest techniques of this type, and show how difficult it is to build a truly secure communication systems.

1 0

Colloquium Speaker David Malan Tues Dec 3, 4:30pm
by Nicole E. Wagenblast 02 Dec '13

02 Dec '13

Pedagogy behind CS50 David Malan , Harvard University Tuesday, December 3, 4:30pm Computer Science 105 CS50 is Harvard University's introductory course for majors and non-majors alike, a one-semester amalgam of courses generally known as CS1 and CS2. In 2007, we set out to alter the course's style and tone to resonate with those "less comfortable" and "more comfortable" alike, albeit without sacrificing the course's historical rigor. We maintained the course's underlying syllabus but revamped every problem set, providing students not only with more direction but context as well. And we augmented the course's support structure. As of 2013, CS50 is Harvard's second-largest course with nearly 700 students, up from 132 in 2006, and those "less comfortable" now compose the course's largest demographic. We present in this talk what we have done and why we have done it. And we offer a glimpse of CS50x, Harvard College's first course to be offered on an even larger scale via edX. BIO David J. Malan is a Senior Lecturer on Computer Science at Harvard University for the School of Engineering and Applied Sciences. He received his A.B., S.M., and Ph.D. in Computer Science from the same in 1999, 2004, and 2007, respectively. His research in graduate school focused primarily on cybersecurity and computer forensics. His more recent publications focus on instructional technologies and pedagogy. He now teaches Harvard College's introductory computer science course, Computer Science 50, otherwise known as CS50, which is available as OpenCourseWare at cs50.tv as well as via edx.org.

1 1

J Lu preFPO
by Melissa M. Lawson 27 Nov '13

27 Nov '13

Jingwan (Cynthia) Lu will present her preFPO on Monday December 9 at 4:30 PM in Room 402. The members of her committee are: Adam Finkelstein, advisor; Szymon Rusinkiewicz and Stephen DiVerdi (Google), readers; Tom Funkhouser and Jianxiong Xiao, nonreaders. Everyone is invited to attend her talk. Her abstract follows below. ----- Original Message ----- Title: Data-driven Digital Drawing and Painting Abstract: Digital artists create evocative drawings and paintings using a tablet and stylus coupled with painting software like Adobe Photoshop or Corel Painter. Several research systems have shown promising improvements in various aspects of the art creation process by targeting specific drawing styles and natural media. Such systems combine carefully hand-crafted procedural rules and computationally expensive, style-specific physical simulations. Nevertheless, untrained users often find it hard to achieve their target style in these systems due to the challenge of controlling and predicting the outcome of their collective strokes. Moreover even trained digital artists are often restricted by the inherent stylistic limitations of these systems. In this thesis, I propose an alternative painting paradigm that allows novices to create artworks closely following an example style of their choice. The style is characterized by a set of example brush strokes. We present three prototype painting systems, all of which follow a common sketching interface, wherein the user's 1D paths are automatically replaced by complex 2D strokes shaped and textured by a small set of exemplars. The three prototype systems HelpingHand, RealBrush and DecoBrush make use of different types of exemplars for distinct styles of artwork. HelpingHand learns from hand gesture data to hallucinate the possible gestures used during drawing, which then drive a chosen virtual brush simulation engine to produce stylistic handwritings or line drawings. RealBrush uses photographed natural media stroke marks to stylize the user's query path, completely obviating the need for physical simulation. It also stylizes the stroke interactions based on captured physical examples. DecoBrush further generalizes the conventional "painting" concept beyond the scope of natural media to allow "painting" structured decorative vector patterns following user-sketched path, which alleviates the difficulty of creating such patterns from scratch in programs like Adobe Illustrator. All three systems address the common challenge of predicting the appearance of user's new query paths being faithful to the style of a small number of chosen exemplars. With the convenience of data sharing on the internet, this data-driven paradigm opens up new opportunities for artists and novices to create original stylistic artworks and to abstract, share, and reproduce their styles more easily and faithfully.

1 0

S Li FPO
by Melissa M. Lawson 27 Nov '13

27 Nov '13

Shi Li will present his FPO on Monday December 9 at 3PM in Room 402. The members of his committee are: Moses Charikar, advisor; Mark Braverman and Julia Chuzhoy (TTI), readers; Bernard Chazelle and Zeev Dvir, nonreaders. A copy of his thesis, "Approximation Algorithms on Network Routing and Facility Location Problems" can be viewed by contacting his advisor. Everyone is invited to attend his talk.

1 0

H Nguyen preFPO
by Melissa M. Lawson 27 Nov '13

27 Nov '13

Huy Nguyen will present his preFPO on Friday December 6 at 4:30 PM in Room 402. The members of his committee are: Moses Charikar, advisor; Bernard Chazelle and Piotr Indyk (MIT), readers; Sanjeev Arora and Mark Braverman, nonreaders. Everyone is invited to attend his talk. His abstract follows below. ----- Original Message ----- Title: Algorithms for Data in High Dimensions In the last decade, the availability of large data sets in high dimensions has been both an opportunity and a challenge for algorithm design. Many new algorithms are invented and yet many challenges remain unconquered. This thesis examines several methods for handling such data efficiently in terms of time and space, their limitations, and new techniques to get beyond those limitations. First, we look at embeddings, a powerful technique to reduce data in high dimensions to low dimensions. Tight or near tight bounds are shown for sparse subspace embeddings, sparse Johnson-Lindenstrauss transforms, and sparse RIP matrices. We also touch upon applications of subspace embeddings in numerical linear algebra and approximating large eigenvalues in both the classical setting and the streaming setting. For instance, the algorithm implied by our bound for subspace embeddings runs in time linear in the sparsity of the input plus an additive term that is optimal unless one can bypass generic matrix multiplication. To conclude the embedding section, we show a near linear lower bound for dimension reduction in L1. This bound is close to optimal as one can reduce the dimensions to linear while approximately preserving all pairwise distances. Even though significant dimension reduction is not possible in L1, fortunately there is a technique called locality sensitive hashing (LSH) allowing for efficient approximate nearest neighbor search in L1. In the second section of the thesis, we give a new construction of LSH for L_p (1<p<2) that is close to the lower bound for LSH. The algorithm is a generalization of previous work for L2 via the uniform convexity property. Lastly, in the final section of the thesis, we overcome the aforementioned lower bound for LSH and design a new algorithm for approximate nearest neighbor search in L1 and L2 that is faster and uses less memory than any LSH. Unlike LSH, the algorithm exploits the structure of the whole point set, and can be viewed as a step toward understanding data-aware methods for nearest neighbor search, which are ubiquitous in practice.

1 0

Y Pritykin preFPO
by Melissa M. Lawson 27 Nov '13

27 Nov '13

Yuri Pritykin will present his preFPO on Friday December 6 at 1:30 PM in Room 402. The members of his committee are: Mona Singh, advisor; Olga Troyanskaya and Bernard Chazelle, readers; Tom Funkhouser and Stanislav Shvartsman (CBE/Genomics), nonreaders. Everyone is invited to attend his talk. His abstract follows below. Title: Intermodular and intramodular proteins in protein interaction networks The availability of large-scale protein-protein interaction networks for numerous organisms provides an opportunity to comprehensively analyze the roles proteins play in the functional organization of the cell. First, we re-examine an influential but controversial characterization of the dynamic modularity of the S. cerevisiae interactome that incorporated gene expression data into network analysis. We analyse the protein-protein interaction networks of five organisms, S. cerevisiae, H. sapiens, D. melanogaster, A. thaliana, and E. coli, and confirm significant and consistent functional and structural differences between hub proteins that are co-expressed with their interacting partners and those that are not, and support the view that the former tend to be intramodular whereas the latter tend to be intermodular. However, we also demonstrate that in each of these organisms, simple topological measures are significantly correlated with the average co-expression of a hub with its partners, and therefore also reflect protein intra- and inter- modularity. Further, cross-interactomic analysis demonstrates that these simple topological characteristics of hub proteins tend to be conserved across organisms. Overall, we give evidence that purely topological features of static interaction networks reflect aspects of the dynamics and modularity of interactomes as well as previous measures incorporating expression data, and are a powerful means for understanding the dynamic roles of hubs in interactomes. Second, we study the role of multifunctional genes (and proteins they encode) in protein interaction networks. Many genes in a genome have more than one function. However, genome-wide analysis of multifunctional genes and an explanation of the mechanisms underlying multifunctionality are still lacking. We leverage functional annotations of genes to extract a group of genes each of which has at least two completely different functions, and distinguish these, the most multifunctional genes, from the remaining genes that are currently known to have a single function. We find biologically significant differences between multifunctional and all other genes on a large scale. We show that multifunctional genes are intermodular and play a central role in the interactome, bridging different functional modules.

1 0

Systems talk on stable multithreading, Junfeng Yang, Monday Nov 25, 3:00pm
by Mike Freedman 25 Nov '13

25 Nov '13

Speaker: Junfeng Yang, Columbia Title: Determinism Is Not Enough: Making Parallel Programs Reliable with Stable Multithreading Date/time: 3-4pm on Monday November 25 Location: CS 402 Abstract: Our accelerating computational demand and the rise of multicore hardware have made parallel programs, especially shared-memory multithreaded programs, increasingly pervasive and critical. Yet, these programs remain extremely difﬁcult to write, test, analyze, debug, and verify. Conventional wisdom has attributed these difﬁculties to nondeterminism (i.e., repeated executions of the same program on the same input may show different behaviors), and researchers have recently dedicated much effort to bringing determinism into multithreading. In this talk, I argue that determinism is not as useful as commonly perceived: it is neither sufﬁcient nor necessary for reliability. We present our view on why multithreaded programs are difﬁcult to get right, describe a promising approach we call stable multithreading to dramatically improve reliability, and summarize our last four years’ research on building and applying stable multithreading systems. Bio: Junfeng Yang's research (http://www.cs.columbia.edu/~junfeng) centers on making reliable and secure systems. He earned his PhD at Stanford, where he created eXplode, a general, lightweight system for effectively finding storage system errors. This work has led to an OSDI '04 best paper, numerous bug fixes to real systems such as the Linux kernel, and a featured article in Linux Weekly news. He worked at Microsoft Research, Silicon Valley from 2007-2008, extending eXplode to check production distributed systems. MoDist, the resultant system, is being transferred to Microsoft product groups. He's now co-directing the Software Systems Lab (ssl.cs.columbia.edu) at Columbia University, where his recent work on making reliable parallel programs---the Tern/Peregrine/Parrot stable multithreading systems--- was featured in CACM, ACM Tech News, The Register, and many other sites. He won Sloan and AFOSR YIP both in 2012; and NSF CAREER in 2011.

1 1

3pm Tue Nov 26 talk on Internet censorship and control in CS 105 (small auditorium)
by Jennifer Rexford 21 Nov '13

21 Nov '13

Speaker: Nick Feamster, Georgia Tech Title: Defending Against Internet Censorship and Control, from Firewalls to Filter Bubbles Date/time: 3-4pm on Tuesday November 26 Location: CS 105 (small auditorium) Abstract: The Internet's promise of open communication and transparency is threatened by both censorship (blocking communication outright) and manipulation (otherwise affecting the performance that a user experiences or the information that a user sees). Censorship is a pervasive threat, with more than 60 countries around the world censoring Internet communications in some form. Unfortunately, many conventional censorship circumvention tools are detectable and can be blocked; in some cases, the use of such software may even be incriminating. Thus, users may need not only to defeat censorship mechanisms but also to hide the fact that they are doing so in the first place. In the first part of the talk, I will describe two systems that achieve this goal, Infranet and Collage. In addition to circumventing censorship firewalls, both Infranet and Collage provide users with the deniability that they are using the censorship circumvention system in the first place. In both systems, we achieve deniability by designing them so that the observable network traffic that they generate is statistically indistinguishable from the user's "normal" traffic patterns if the tools were not in use. Infranet achieves deniability by hiding a user's requests for Web content in other Web traffic that resembles a user's typical browsing pattern. Collage achieves deniability by hiding content in user-generated content sites and encoding a user's request for the content in a sequence of operations that a given user is likely to perform. In recent years, however, circumvention alone is insufficient: sophisticated organizations can also control users by manipulating network traffic. Manipulation can take many forms, from degrading performance to such an extent that a user does not want to use the service, to using Internet communication to generate propaganda (e.g., via social media), to attacking personalization algorithms to affect the results that a user sees in response to a search query. In the second part of the talk, I will describe several manipulation attacks that we have studied and inference techniques that we have developed to detect them. I will conclude with our ongoing efforts to tackle open challenges in this area, such as the deceptively challenging problem of confirming the existence of various forms of censorship in the first place. Bio: Nick Feamster is an associate professor in the College of Computing at Georgia Tech. He received his Ph.D. in Computer science from MIT in 2005, and his S.B. and M.Eng. degrees in Electrical Engineering and Computer Science from MIT in 2000 and 2001, respectively. His research focuses on many aspects of computer networking and networked systems, with a focus on network operations, network security, and censorship-resistant communication systems. In December 2008, he received the Presidential Early Career Award for Scientists and Engineers (PECASE) for his contributions to cybersecurity, notably spam filtering. His honors include the Technology Review 35 "Top Young Innovators Under 35" award, the ACM SIGCOMM Rising Star Award, a Sloan Research Fellowship, the NSF CAREER award, the IBM Faculty Fellowship, the IRTF Applied Networking Research Prize, and award papers at the SIGCOMM Internet Measurement Conference (measuring Web performance bottlenecks), SIGCOMM (network-level behavior of spammers), the NSDI conference (fault detection in router configuration), Usenix Security (circumventing web censorship using Infranet), and Usenix Security (web cookie analysis).

1 0