- talks - lists.cs.princeton.edu

Colloquium Speaker Tom Kwiatkowski TODAY at 1:30pm
by Emily C. Lawrence 03 Dec '19

03 Dec '19

CS Colloquium Speaker Tom Kwiatkowski, Google Tuesday, December 3 - 1:30PM Computer Science - Room 104 Host: Karthik Narasimhan [ https://www.cs.princeton.edu/events/25905 | https://www.cs.princeton.edu/events/25905 ] New Challenges in Question Answering: Natural Questions and Going Beyond Word Matching Abstract: Recently, learned deep models have surpassed human performance on a number of question answering benchmarks such as SQuAD . However, these models resort to simple word matching and answer typing heuristics, and they are easily fooled. In this talk, I will present two different lines of work that aim to take us beyond the current status quo in question answering, and push us towards more robust representations of semantics, pragmatics, and knowledge. First, I will present Natural Questions (NQ), a new question answering benchmark from Google. NQ contains real user questions, which require an understanding of the questioner's intent. NQ also requires systems to read entire Wikipedia pages to decide whether they fully answer a question. This is much harder than finding an answer given the knowledge that one is present. I will convince you that the question 'when was the last time a hurricane hit Massachusetts?' is under-specified with many reasonable answers, and I will tell you how we developed robust evaluation metrics to deal with this ambiguity. In the second part of the talk I will present a complementary method of challenging today's question answering systems by removing access to evidence documents at inference time. Instead of building joint representations of questions and documents, we perform ahead of inference time reading and retrieve answers via fast maximum inner product search. I will show that this leads to large gains in accuracy and speed when finding answers in very large corpora. I will also show some preliminary results that show how our methods can be used to aggregate information from multiple diverse documents.

1 0

Noah Apthorpe will present his Pre FPO "Network Privacy and User Protection in the Internet of Things" on Wednesday, December 4, 2019 at 10:30am in CS 402
by Nicki Mahler 02 Dec '19

02 Dec '19

Noah Apthorpe will present his Pre FPO, "Network Privacy and User Protection in the Internet of Things" on Wednesday, December 4, 2019 at 10:30am in CS 402. The members of his committee are as follows: Nick Feamster (adviser), Examiners: Ed Felten and Arvind Narayanan; Readers: Jennifer Rexford and Marshini Chetty Everyone is invited to attend his talk. The talk abstract follows below. In this talk, I will present my research combining networks, human-computer interaction, and machine learning to study privacy repercussions of Internet of things (IoT) devices and associated online services. This interdisciplinary approach has provided a unique perspective for measuring privacy issues at scale, protecting users from privacy vulnerabilities, and understanding how privacy norms interact with data collection and regulation. First, I will demonstrate that passive network observers can easily infer private in-home user activities from IoT network traffic metadata. This is possible because most user interactions with home IoT devices cause noticeable changes in network traffic rates that are detectable with thresholds or machine learning methods. The limited functionality of these devices makes it easy to associate these traffic rate changes with specific user activities. Fixed distribution traffic shaping can efficiently protect IoT devices that tolerate long network latencies or that have relatively constant traffic rates, but devices with more complicated network activity require more sophisticated protections. I will discuss a zero-latency traffic shaping algorithm, “Stochastic Traffic Padding,” that mimics user activity patterns, provides probabilistic limits on adversary inference accuracy, and allows for tuning of privacy protection versus bandwidth overhead. Next, I will present a survey instrument based on the theory of contextual integrity that allows researchers to measure individual and societal privacy norms at scale. My collaborators and I first applied this survey method to discover home IoT privacy norms of 1,731 U.S. adults, providing recommendations for IoT device manufacturers, regulators, and consumer advocates. We then surveyed 195 U.S. parents, showing that data handling requirements from the Children’s Online Privacy Protection Act (COPPA) generally align with parents’ privacy expectations for IoT toys. However, observed variations in the perceived acceptability of data collection across specific toys, information types, and other conditions emphasize the importance of detailed contextual factors to privacy norms and suggest potential improvements to COPPA. I will conclude with a summary of my ongoing research and a timeline for the completion of my degree.

1 0

Reminder: Colloquium Speaker Tom Kwiatkowski, tomorrow at 1:30pm
by Emily C. Lawrence 02 Dec '19

02 Dec '19

CS Colloquium Speaker Tom Kwiatkowski, Google Tuesday, December 3 - 1:30PM Computer Science - Room 104 Host: Karthik Narasimhan [ https://www.cs.princeton.edu/events/25905 | https://www.cs.princeton.edu/events/25905 ] New Challenges in Question Answering: Natural Questions and Going Beyond Word Matching Abstract: Recently, learned deep models have surpassed human performance on a number of question answering benchmarks such as SQuAD . However, these models resort to simple word matching and answer typing heuristics, and they are easily fooled. In this talk, I will present two different lines of work that aim to take us beyond the current status quo in question answering, and push us towards more robust representations of semantics, pragmatics, and knowledge. First, I will present Natural Questions (NQ), a new question answering benchmark from Google. NQ contains real user questions, which require an understanding of the questioner's intent. NQ also requires systems to read entire Wikipedia pages to decide whether they fully answer a question. This is much harder than finding an answer given the knowledge that one is present. I will convince you that the question 'when was the last time a hurricane hit Massachusetts?' is under-specified with many reasonable answers, and I will tell you how we developed robust evaluation metrics to deal with this ambiguity. In the second part of the talk I will present a complementary method of challenging today's question answering systems by removing access to evidence documents at inference time. Instead of building joint representations of questions and documents, we perform ahead of inference time reading and retrieve answers via fast maximum inner product search. I will show that this leads to large gains in accuracy and speed when finding answers in very large corpora. I will also show some preliminary results that show how our methods can be used to aggregate information from multiple diverse documents.

1 0

Laura Roberts will present her Pre FPO "Prevention and Detection of Internet Surveillance" on Tuesday, December 3, 2019 at 2:30pm in CS 402
by Nicki Mahler 27 Nov '19

27 Nov '19

Laura Roberts will present her Pre FPO on Tuesday, December 3, 2019 at 2:30pm in CS 402. The members of her committee are as follows: Ed Felten (adviser); Readers: Prateek Mittal and Jonathan Mayer; Examiners: Jennifer Rexford and Arvind Narayanan Everyone is invited to attend her talk. The talk title and abstract follow below. Title: Prevention and Detection of Internet Surveillance If you believe that privacy is important in a free and democratic society, then you are surely worried about the enormous surveillance system that is the Internet. In this talk, I present practical ways to prevent being watched while using the Internet by plugging holes in Tor and a practical way to watch who’s watching you. First, we investigate the effect of DNS on Tor’s anonymity. Previous work had not analyzed DNS’s role in deanonymizing Tor users. We discovered that DNS exposes Tor users’ behavior to more adversaries than previously thought, we discovered that Google gets to learn a lot about what websites Tor users are visiting via DNS, we performed simulations at Internet-scale in order to understand how our proof-of-concept attacks could affect real people, and finally, we provided recommendations to Tor operators on how they should be handling DNS in Tor. Next, we analyzed millions of RSA keys in Tor relays in search of anomalies. Previous work has shown how weak cryptographic keys can compromise users’ safety by allowing adversaries to compute the all-important private key. We indeed found anomalous keys and demonstrated that they were indicative of suspicious behavior. Our results led to The Tor Project’s taking action to develop tools that catch these anomalous keys. Finally, Internet users and providers of services don't know who's watching their Internet traffic. We desire a way to detect who is monitoring Internet traffic and to determine where that traffic might be monitored. Thus, we present NOISE, the Nonce Observatory for Inverse Surveillance of Eavesdroppers, a method and system to detect eavesdroppers remotely that involves disseminating nonces--unique, pseudorandom values--in traffic and seeing if they are observed elsewhere, indicating they must have been "eavesdropped" and acted upon unexpectedly. We take advantage of IPv6's large address space and embed 64-bit nonces innocuously into IPv6 addresses. We disseminate these "nonced" addresses via Internet-wide, traceroute-like active measurement campaigns to approx. 15.2M targets as unique source addresses in each outbound probe. We monitor for subsequent interest in these nonces via packet capture on our system's infrastructure and inspection of system logs, email, and a third-party passive DNS database. Disseminating ~1.2 billion nonces, we detect eavesdropping more than 200k times, involving 268 networks, for probes destined for 437 networks. These observations reveal: (a) data collection for security incident handling, (b) traffic information being shared with third parties, and (c) man-in-the-middle monitoring in one of the world's largest commercial peering exchanges.

1 0

Colloquium Speaker Tom Kwiatkowski, Tues. Dec 3
by Emily C. Lawrence 27 Nov '19

27 Nov '19

CS Colloquium Speaker Tom Kwiatkowski, Google Tuesday, December 3 - 1:30PM Computer Science - Room 104 Host: Karthik Narasimhan [ https://www.cs.princeton.edu/events/25905 | https://www.cs.princeton.edu/events/25905 ] New Challenges in Question Answering: Natural Questions and Going Beyond Word Matching Abstract: Recently, learned deep models have surpassed human performance on a number of question answering benchmarks such as SQuAD . However, these models resort to simple word matching and answer typing heuristics, and they are easily fooled. In this talk, I will present two different lines of work that aim to take us beyond the current status quo in question answering, and push us towards more robust representations of semantics, pragmatics, and knowledge. First, I will present Natural Questions (NQ), a new question answering benchmark from Google. NQ contains real user questions, which require an understanding of the questioner's intent. NQ also requires systems to read entire Wikipedia pages to decide whether they fully answer a question. This is much harder than finding an answer given the knowledge that one is present. I will convince you that the question 'when was the last time a hurricane hit Massachusetts?' is under-specified with many reasonable answers, and I will tell you how we developed robust evaluation metrics to deal with this ambiguity. In the second part of the talk I will present a complementary method of challenging today's question answering systems by removing access to evidence documents at inference time. Instead of building joint representations of questions and documents, we perform ahead of inference time reading and retrieve answers via fast maximum inner product search. I will show that this leads to large gains in accuracy and speed when finding answers in very large corpora. I will also show some preliminary results that show how our methods can be used to aggregate information from multiple diverse documents.

1 0

Colloquium Speaker Jason Weston TODAY at 1:30pm
by Emily C. Lawrence 26 Nov '19

26 Nov '19

CS Colloquium Speaker Jason Weston, Facebook Tuesday, November 26 - 1:30PM Computer Science - Room 104 Host: Danqi Chen [ https://www.cs.princeton.edu/events/25903 | https://www.cs.princeton.edu/events/25903 ] Putting together the threads of conversational AI Maybe we don't have enough threads yet to knit together the whole, but let's try anyway! We present our view of what is necessary for conversational AI, and the pieces we have worked on so far to get there. In particular: software ( ParlAI , a unified platform for dialogue research), various neural architectures for memory, reasoning, retrieval and generation, and interactive learning, tasks for employing personality (Persona-Chat), knowledge (Wizard of Wikipedia) and perception (Image-Chat), evaluation studies & techniques (Acute- Eval ), and a recent competition (ConvAI2). We'll try not to be too pessimistic (!), but we'll also discuss how far we still have to go. Bio: Jason Weston is a research scientist at Facebook, NY and a Visiting Research Professor at NYU. He earned his PhD in machine learning at Royal Holloway, University of London and at AT&T Research in Red Bank, NJ (advisors: Alex Gammerman , Volodya Vovk and Vladimir Vapnik ) in 2000. From 2000 to 2001, he was a researcher at Biowulf technologies. From 2002 to 2003 he was a research scientist at the Max Planck Institute for Biological Cybernetics, Tuebingen , Germany. From 2003 to 2009 he was a research staff member at NEC Labs America, Princeton. From 2009 to 2014 he was a research scientist at Google, NY. His interests lie in statistical machine learning, with a focus on reasoning, memory, perception, interaction and communication. Jason has published over 100 papers, including best paper awards at ICML and ECML, and a Test of Time Award for his work "A Unified Architecture for Natural Language Processing: Deep Neural Networks with Multitask Learning", ICML 2008 (with Ronan Collobert ). He was part of the YouTube team that won a National Academy of Television Arts & Sciences Emmy Award for Technology and Engineering for Personalized Recommendation Engines for Video Discovery. He was listed as the 16th most influential machine learning scholar at AMiner and one of the top 50 authors in Computer Science in Science.

1 0

Distinguished CS Colloquium Speaker Subhash Khot, Monday Dec 9
by Emily C. Lawrence 25 Nov '19

25 Nov '19

Distinguished CS Colloquium Speaker Subhash Khot, New York University Monday, December 9 - 12:30pm Computer Science - Room 105 Host: Ran Raz [ https://www.cs.princeton.edu/events/25898 | https://www.cs.princeton.edu/events/25898 ] Hardness of Approximation: From the PCP Theorem to the 2-to-2 Games Theorem Computer scientists firmly believe that no algorithm can efficiently compute optimal solutions to a class of problems called NP-hard problems. For many NP-hard problems, even computing an approximately optimal solution remains NP-hard. This phenomenon, known as the hardness of approximation, has numerous connections to proof checking, optimization, combinatorics , discrete Fourier analysis, and geometry. The talk will provide an overview of these connections. It will address why graph coloring is a computationally hard problem, how it is possible to check a proof without even looking at it, why computer scientists love the majority vote, and whether a shape exists that looks spherical as well as cubical. It will explain how all this fits into a 30-year research program starting with the foundational Probabilistically Checkable Proofs (PCP) Theorem and leading to the recent 2-to-2 Games Theorem. Bio: Subhash Khot is a professor of computer science at the Courant Institute of Mathematical Sciences at New York University. His prior affiliations include Princeton University (PhD), Institute for Advanced Study (member), Georgia Tech (assistant professor) and University of Chicago (visiting professor). His research centers on computational complexity and its connections to geometry and analysis. He is a recipient of the National Science Foundation’s Alan T. Waterman Award, the International Mathematical Union’s Nevanlinna Prize, the Simons Investigator Award, a MacArthur Fellowship, and is a Fellow of the Royal Society.

1 0

Reminder: Colloquium Speaker Jason Weston, tomorrow at 1:30pm
by Emily C. Lawrence 25 Nov '19

25 Nov '19

CS Colloquium Speaker Jason Weston, Facebook Tuesday, November 26 - 1:30PM Computer Science - Room 104 Host: Danqi Chen [ https://www.cs.princeton.edu/events/25903 | https://www.cs.princeton.edu/events/25903 ] Putting together the threads of conversational AI Maybe we don't have enough threads yet to knit together the whole, but let's try anyway! We present our view of what is necessary for conversational AI, and the pieces we have worked on so far to get there. In particular: software ( ParlAI , a unified platform for dialogue research), various neural architectures for memory, reasoning, retrieval and generation, and interactive learning, tasks for employing personality (Persona-Chat), knowledge (Wizard of Wikipedia) and perception (Image-Chat), evaluation studies & techniques (Acute- Eval ), and a recent competition (ConvAI2). We'll try not to be too pessimistic (!), but we'll also discuss how far we still have to go. Bio: Jason Weston is a research scientist at Facebook, NY and a Visiting Research Professor at NYU. He earned his PhD in machine learning at Royal Holloway, University of London and at AT&T Research in Red Bank, NJ (advisors: Alex Gammerman , Volodya Vovk and Vladimir Vapnik ) in 2000. From 2000 to 2001, he was a researcher at Biowulf technologies. From 2002 to 2003 he was a research scientist at the Max Planck Institute for Biological Cybernetics, Tuebingen , Germany. From 2003 to 2009 he was a research staff member at NEC Labs America, Princeton. From 2009 to 2014 he was a research scientist at Google, NY. His interests lie in statistical machine learning, with a focus on reasoning, memory, perception, interaction and communication. Jason has published over 100 papers, including best paper awards at ICML and ECML, and a Test of Time Award for his work "A Unified Architecture for Natural Language Processing: Deep Neural Networks with Multitask Learning", ICML 2008 (with Ronan Collobert ). He was part of the YouTube team that won a National Academy of Television Arts & Sciences Emmy Award for Technology and Engineering for Personalized Recommendation Engines for Video Discovery. He was listed as the 16th most influential machine learning scholar at AMiner and one of the top 50 authors in Computer Science in Science.

1 0

Colloquium speaker Eleazar Eskin, TODAY at 12:30PM
by Emily C. Lawrence 22 Nov '19

22 Nov '19

CS Colloquium Speaker Eleazar Eskin , University of California, Los Angeles Friday, November 22 - 12:30pm Computer Science - Room 105 Host: Ben Raphael [ https://www.cs.princeton.edu/events/25902 | https://www.cs.princeton.edu/events/25902 ] Leveraging Allelic Heterogeneity in Association Studies Over 100,000 variants have been implicated in human traits through genome-wide association studies (GWAS). Virtually all reported GWAS identify associations by measuring the correlation between a single variant and a phenotype of interest. Recently, several studies reported that at many risk loci, there may exist multiple causal variants, a phenomenon referred to as allelic heterogeneity. Naturally, for a locus with multiple causal variants with small effect sizes, the standard association test is underpowered to detect the associations. Alternatively, an approach considering effects of multiple variants simultaneously may increase statistical power. Counterintuitively, most approaches that consider multiple variants in association studies find fewer associations than the single SNP association test. This is due to the fact that most multiple variant methods assume a structure of allelic heterogeneity which is very different from what is observed in genetic studies. In this work, we propose a new statistical method, Model-based Association test Reflecting causal Status (MARS), that tries to find an association between variants in risk loci and a phenotype, considering multiple variants at each locus. One of the main advantages of MARS is that it only requires the existing summary statistics to detect associated risk loci. Thus, MARS is applicable to any association study with summary statistics, even though individual level data is not available for the study. Utilizing extensive simulated data sets, we show that MARS increases the power of detecting true associated risk loci compared to previous approaches that consider multiple variants, while robustly controls the type I error. Applied to data of 44 tissues provided by the Genotype- Tissue Expression ( GTEx ) consortium, we show that MARS identifies more eGenes compared to previous approaches in most of the tissues; e.g. MARS identified 16% more eGenes than the ones reported by the GTEx consortium. Moreover, applied to Northern Finland Birth Cohort (NFBC) data, we demonstrate that MARS effectively identifies association loci with improved power (56% of more loci found by MARS) in GWAS studies compared to the standard association test. Bio: Dr. Eleazar Eskin serves as the inaugural chair for the UCLA Department of Computational Medicine. Fascinated by the intersection of computer science and biology, Dr. Eskin is researching and developing computational methods for the analysis of genetic variations in human disease. There are millions of variants in the genome and identifying the variants involved in disease often requires tens of thousands of patient samples. In order to analyze these tremendously large datasets, Dr. Eskin and his team are solving challenging computational problems and developing new computational techniques. He received his PhD in computer science from Columbia University. A recipient of the Alfred P. Sloan Foundation Research Fellowship, Dr. Eskin’s work is supported by the National Science Foundation and the National Institutes of Health.

1 0

Speaker David Hay TODAY at 11:00AM
by Emily C. Lawrence 22 Nov '19

22 Nov '19

CS Speaker David Hay, Hebrew University Friday, November 22 - 11:00AM Computer Science - Room 105 Host: Jennifer Rexford [ https://www.cs.princeton.edu/events/25900 | https://www.cs.princeton.edu/events/25900 ] Network Level IoT Security Computer networks have undergone and continue to experience a major transformation, whereby billions of low-cost devices are being connected to the network to provide additional functionality and better user experience. Unlike traditional network devices, these devices, collectively known as the ``Internet of Things'' ( IoT ), typically have very limited computational, memory, and power resources. These IoT devices became a major security concerns, both due to human factors and to technical challenges in deploying security mechanisms on devices with low resources. The number and diversity of IoT devices creates a huge attack surface that is often exploited by attackers to launch large-scale attacks, sometimes exploiting well-known vulnerabilities. This talk will highlight the security concerns of IoT devices from a networking perspective and explore how to secure IoT devices using whitelists, in which communication between a device and an endpoint is prohibited unless that endpoint appears in the corresponding whitelist. Finally, we will discussed deployment options for such a solution (namely, within the internet gateway, as virtual network function within the ISP network, or a combination of the two). Bio: David Hay is an Associate Professor with the Rachel and Selim Benin School of Computer Science and Engineering, Hebrew University, Jerusalem, Israel. He received the B.A. (summa cum laude) and Ph.D. degrees in computer science from the Technion —Israel Institute of Technology, Haifa, Israel, in 2001 and 2007, respectively. In addition, he was with IBM Haifa Research Labs, Haifa, Israel; Cisco Systems, San Jose, CA, USA; the Electronic Department, Politecnico di Torino, Turin, Italy; and the Electrical Engineering Department with Columbia University, New York, NY, USA. In 2010, he co-founded (with Prof. Brembler -Barr) the DEEPNESS lab, focusing on deep packet inspection in next-generation network devices. He has served as a technical program committee member of numerous networking conferences, and since 2018 serves as en editor of ACM/IEEE Transactions on Networking. His research interests are in computer networks—in particular, network algorithmics , packet classification, deep packet inspection, network survivability and resilience, software-defined networking, network-function virtualization, and various aspects of network security.

1 0