May 2017 - talks - lists.cs.princeton.edu

Jordan Ash will present his general exam on Tuesday, May 9th, 2017 at 10am in CS 401. (updated reading list)
by Nicki Gotsis 05 May '17

05 May '17

Jordan Ash will present his general exam on Tuesday, May 9th, 2017 at 10am in CS 401. The members of his committee are: Barbara Engelhardt (adviser), Sebastian Seung, and Tom Funkhouser. Everyone is invited to attend his talk, and those faculty wishing to remain for the oral exam following are welcome to do so. His abstract and reading list follow below. Abstract: Domain adaptation addresses the problem created when training data is generated by a so-called source distribution, but test data is generated by a significantly different target distribution. In this work, we present approximate label matching (ALM), a new unsupervised domain adaptation technique that creates and leverages a rough labeling on the test samples, then uses these noisy labels to learn a transformation that aligns the source and target samples. We show that the transformation estimated by ALM has favorable properties compared to transformations estimated by other methods, which do not use any kind of target labeling. Our model is regularized by requiring that a classifier trained to discriminate source from transformed target samples cannot distinguish between the two. We experiment with ALM on simulated and real data, and show that it outperforms techniques commonly used in the field. Keywords: Generative Adversarial Networks, Domain Adaptation, Deep Learning Reading List: 1. Rumelhart, David E., Geoffrey E. Hinton, and Ronald J. Williams. "Learning representations by back-propagating errors." Cognitive modeling (1988) 2. LeCun, Yann, et al. "Gradient-based learning applied to document recognition." Proceedings of the IEEE 86.11 (1998): 2278-2324. 3. LeCun, Yann, Yoshua Bengio, and Geoffrey Hinton. "Deep learning." Nature 521.7553 (2015): 436-444. 4. Ganin, Yaroslav, et al. "Domain-adversarial training of neural networks." Journal of Machine Learning Research 17.59 (2016): 1-35. 5. Goodfellow, Ian, et al. "Generative adversarial nets." Advances in neural information processing systems. 2014. 6. Denton, Emily L., Soumith Chintala, and Rob Fergus. "Deep Generative Image Models using a Laplacian Pyramid of Adversarial Networks." Advances in neural information processing systems. 2015. 7. Radford, Alec, Luke Metz, and Soumith Chintala. "Unsupervised representation learning with deep convolutional generative adversarial networks." arXiv preprint arXiv:1511.06434 (2015) 8. Gardner, Jacob R., et al. "Deep manifold traversal: Changing labels with convolutional features." arXiv preprint arXiv:1511.06421 (2015). 9. Liu, Ming-Yu, and Oncel Tuzel. "Coupled generative adversarial networks." Advances in Neural Information Processing Systems. 2016. 10. Mathieu, Michael, Camille Couprie, and Yann LeCun. "Deep multi-scale video prediction beyond mean square error." arXiv preprint arXiv:1511.05440 (2015). 11. Lamb, Alex M., et al. "Professor forcing: A new algorithm for training recurrent networks." Advances In Neural Information Processing Systems. 2016.

1 0

Noah Apthorpe will present his general exam on Friday, May 12, 2017 in CS 301 at 3:30pm
by Nicki Gotsis 05 May '17

05 May '17

Noah Apthorpe will present his general exam on Friday, May 12, 2017 in CS 301 at 3:30pm The members of his committee are: Nick Feamster, Jennifer Rexford, and Ed Felten Everyone is invited to attend his talk, and those faculty wishing to remain for the oral exam following are welcome to do so. His abstract and reading list follow below. BQ_BEGIN Privacy risks of s mart home Internet-of-things traffic analysis and low-cost solutions The growing market for smart home IoT devices promises new conveniences for consumers while presenting novel challenges for preserving privacy within the home. Many smart home devices have always-on sensors that capture users’ offline activities in their living spaces and transmit information about these activities on the Internet. The recent reversal of the FCC broadband privacy rules will allow Internet Service Providers (ISPs) to continue collecting, and potentially selling, consumer data obtained by network traffic analysis. Here, we demonstrate that an ISP can infer privacy sensitive in-home activities using IPFIX (NetFlow) records or packet captures from smart homes containing commercially-available IoT devices even when the devices use SSL encryption . We show that well-studied traffic shaping techniques for preventing traffic correlation attacks on anonymity networks are also effective at preventing activity inference from smart home traffic. However, these traffic shaping techniques trade off latency and cover traffic bandwidth, which can result in unacceptable device performance or expensive pay-per-use Internet access fees. This tradeoff is required because these techniques assume an adversary with packet captures that include all packet sizes, send times, and inter-packet intervals. In reality, ISPs predominantly collect IPFIX or NetFlow records from routers and switches rather than packet captures. We propose and evaluate a technique for preventing smart home activity inference from sampled or unsampled NetFlow records that does not effect network latency or device performance and requires minimal bandwidth overhead. This technique complements increased interest in virtual private networks (VPNs) since the reversal of the FCC broadband privacy rules and is deployable on smart home devices, third-party IoT hubs, or gateway routers. Note: This work is a continuation of the workshop paper " A Smart Home is No Castle: P rivacy Vulnerabilities of Encrypted IoT Traffic " by Apthorpe, et al. Textbook Computer Networks (5th Edition) by Peterson & Davie Papers Shmatikov, Vitaly, and Ming-Hsiu Wang. "Timing analysis in low-latency mix networks: Attacks and defenses." In European Symposium on Research in Computer Security , pp. 18-33. Springer Berlin Heidelberg, 2006. ( PDF ) Fu, Xinwen, Bryan Graham, Riccardo Bettati, Wei Zhao, and Dong Xuan. "Analytical and empirical analysis of countermeasures to traffic analysis attacks." In Parallel Processing, 2003. Proceedings. 2003 International Conference on , pp. 483-492. IEEE, 2003. ( PDF ) Wang, Wei, Mehul Motani, and Vikram Srinivasan. "Dependent link padding algorithms for low latency anonymity systems." In Proceedings of the 15th ACM conference on Computer and communications security , pp. 323-332. ACM, 2008. ( PDF ) Dyer, Kevin P., Scott E. Coull, Thomas Ristenpart, and Thomas Shrimpton. "Peek-a-boo, I still see you: Why efficient traffic analysis countermeasures fail." In Security and Privacy (SP), 2012 IEEE Symposium on , pp. 332-346. IEEE, 2012. ( PDF ) Srinivasan, Vijay, John Stankovic, and Kamin Whitehouse. "Protecting your daily in-home activity information from a wireless snooping attack." In Proceedings of the 10th international conference on Ubiquitous computing , pp. 202-211. ACM, 2008. ( PDF ) Back, Adam, Ulf Möller, and Anton Stiglic. "Traffic analysis attacks and trade-offs in anonymity providing systems." In International Workshop on Information Hiding , pp. 245-257. Springer Berlin Heidelberg, 2001. ( PDF ) Verde, Nino Vincenzo, Giuseppe Ateniese, Emanuele Gabrielli, Luigi Vincenzo Mancini, and Angelo Spognardi. "No NAT'd user left behind: Fingerprinting users behind nat from Netflow records alone." In Distributed Computing Systems (ICDCS), 2014 IEEE 34th International Conference on , pp. 218-227. IEEE, 2014. ( PDF ) Bellovin, Steven M. "A technique for counting NATted hosts." In Proceedings of the 2nd ACM SIGCOMM Workshop on Internet measurment , pp. 267-272. ACM, 2002. ( PDF ) Kohno, Tadayoshi, Andre Broido, and Kimberly C. Claffy. "Remote physical device fingerprinting." IEEE Transactions on Dependable and Secure Computing 2, no. 2 (2005): 93-108. ( PDF ) Rieke, Aaron, David Robinson, and Harlan Yu. "What ISPs can see." Upturn. 2016. ( PDF ) Nissenbaum, Helen. "Privacy as contextual integrity." Wash. L. Rev. 79 (2004): 119. ( PDF ) BQ_END

1 0

TODAY at 4:30 pm -- EE Seminar - May 3, 2017, 4:30 pm, E-Quad B205 - Dr. Sukarno Mertoguno
by Nicole E. Wagenblast 03 May '17

03 May '17

Department of Electrical Engineering Seminar Speaker: Dr. Sukarno Mertoguno The Office of Naval Research Title: Research in Cyber Security & Complex Software at ONR code 311 Date: Wednesday, May 3, 2017 Time: 4:30 pm Room: E -Quad, Room B205 Host: Prof. Prateek Mittal Abstract: This talk will discuss a quick overview of current research topics for Cyber Security & Complex Software at ONR code 311. Emphasis will be given to our Learn-2-Reason concept, a hybrid formal-statistical decision making system, and topics related to our current and incoming BAAs on specialization for software and protocols, toward achieving more efficient and more secure cyber systems. Bio: Dr. J. Sukarno Mertoguno manages basic and applied science research in cyber security and complex software for The Office of Naval Research (ONR). Before joining ONR he worked as a system & chip architect and an entrepreneur in the Silicon Valley, where he has worked on various chips and systems, such as embedded processors, switching fabric, network processors, and various other hardware accelerators, including TCP/IP, NFS, mobile anti-malware, etc. He received a Ph.D. in electrical engineering from SUNY-Binghamton. He also has background in Theoretical Physics. Save

1 0

Rafael Oliveira will present his Pre FPO, "Polynomial identity testing: derandomization results and applications" on Monday, May 8, 2017 at 2pm in CS 302.
by Nicki Gotsis 02 May '17

02 May '17

Rafael Oliveira will present his Pre FPO, " Polynomial identity testing: derandomization results and applications" on Monday, May 8, 2017 at 2pm in CS 302. The members of his committee are: Zeev Dvir (adviser/reader), Gillat Kol (reader), Mark Braverman (examiner), and Ran Raz (examiner) Everyone is invited to attend his talk. The talk title and abstract follow below: Title: Polynomial identity testing: derandomization results and applications Abstract: The purpose of algebraic complexity theory is to provide a framework for understanding the powers and limitations of computation when the resources used and the computations performed are of an algebraic nature. Many important computational problems in computer science, mathematics and physics can be cast in an algebraic way, and understanding the computational complexity of the algebraic structures arising from such formulations is one of the chief goals of the theory. In this thesis, we study two fundamental problems in algebraic complexity theory: the derandomization of the polynomial identity testing problem (and of its non-commutative variant) and the parallel complexity of factoring polynomials. We will show connections between these two problems and how these problems can be used to solve computational questions in invariant theory, non-commutative algebra, quantum physics and combinatorics.

1 0

Sophie Qiu will present her MSE talk, "Streaming data analytics and visualization: An illustrative example" on Monday, May 08, 2017 at 11am in CS 301
by Nicki Gotsis 02 May '17

02 May '17

Sophie Qiu will present her MSE talk, "Streaming data analytics and visualization: An illustrative example" on Monday, May 08, 2017 at 11am in CS 301. The members of her committee are Jen Rexford (adviser) and Walter Willinger (Niksun). Everyone is invited to attend her talk. Her abstract follows below. The emergence of streaming data or “data in motion” has motivated the development of new “streaming” algorithms that provide up-to-date answers to continuous queries; that is, queries that are issued once and then run continuously as new data streams in. For example, in the context of network traffic management, continuous queries over streaming Netflow data may be used to detect anomalies in the network as they happen (e.g., performance degradation, onset of an attack). The main objective of this thesis is to demonstrate with an illustrative example how the continuous stream of answers produced by existing streaming algorithms can be visualized in an effective and meaningful manner. In particular, such visualization efforts should be able to capture the dynamics that is inherent in the answers that result from such continuous queries. Our illustrative example concerns the class of frequent itemset mining algorithms for streaming data, a generalization of the well-know algorithms for finding frequent items (e.g., top-k) in streaming data. Using the output of these streaming algorithms as input, our proposed visualization method c ombines two previously-considered types of flow diagrams (i.e., sankey and alluvial diagrams) to depict the temporal evolution of frequent itemsets. We apply our visualization method to real-world streaming data (e.g., unsampled Netflow data from a stub network, DNS records from a university network) and illustrate its basic features and capabilities.

1 0

Jordan Ash will present his general exam on Tuesday, May 9th, 2017 at 10am in CS 401.
by Nicki Gotsis 02 May '17

02 May '17

Jordan Ash will present his general exam on Tuesday, May 9th, 2017 at 10am in CS 401. The members of his committee are: Barbara Engelhardt (adviser), Sebastian Seung, and Tom Funkhouser. Everyone is invited to attend his talk, and those faculty wishing to remain for the oral exam following are welcome to do so. His abstract and reading list follow below. Abstract: Domain adaptation addresses the problem created when training data is generated by a so-called source distribution, but test data is generated by a significantly different target distribution. In this work, we present approximate label matching (ALM), a new unsupervised domain adaptation technique that creates and leverages a rough labeling on the test samples, then uses these noisy labels to learn a transformation that aligns the source and target samples. We show that the transformation estimated by ALM has favorable properties compared to transformations estimated by other methods, which do not use any kind of target labeling. Our model is regularized by requiring that a classifier trained to discriminate source from transformed target samples cannot distinguish between the two. We experiment with ALM on simulated and real data, and show that it outperforms techniques commonly used in the field. Keywords: Generative Adversarial Networks, Domain Adaptation, Deep Learning Reading List: 1. Rumelhart, David E., Geoffrey E. Hinton, and Ronald J. Williams. "Learning representations by back-propagating errors." Cognitive modeling (1988) 2. LeCun, Yann, et al. "Gradient-based learning applied to document recognition." Proceedings of the IEEE 86.11 (1998): 2278-2324. 3. LeCun, Yann, Yoshua Bengio, and Geoffrey Hinton. "Deep learning." Nature 521.7553 (2015): 436-444. 4. Ganin, Yaroslav, et al. "Domain-adversarial training of neural networks." Journal of Machine Learning Research 17.59 (2016): 1-35. 5. Goodfellow, Ian, et al. "Generative adversarial nets." Advances in neural information processing systems. 2014. 6. Denton, Emily L., Soumith Chintala, and Rob Fergus. "Deep Generative Image Models using a Laplacian Pyramid of Adversarial Networks." Advances in neural information processing systems. 2015. 7. Radford, Alec, Luke Metz, and Soumith Chintala. "Unsupervised representation learning with deep convolutional generative adversarial networks." arXiv preprint arXiv:1511.06434 (2015). 8. Reed, Scott, et al. "Learning to disentangle factors of variation with manifold interaction." Proceedings of the 31st International Conference on Machine Learning (ICML-14). 2014. 9. Liu, Ming-Yu, and Oncel Tuzel. "Coupled generative adversarial networks." Advances in Neural Information Processing Systems. 2016. 10. Mathieu, Michael, Camille Couprie, and Yann LeCun. "Deep multi-scale video prediction beyond mean square error." arXiv preprint arXiv:1511.05440 (2015). 11. Lamb, Alex M., et al. "Professor forcing: A new algorithm for training recurrent networks." Advances In Neural Information Processing Systems. 2016.

1 0

Ari Seff will present his general exam on Monday, May 8th at 9am in CS 302
by Nicki Gotsis 01 May '17

01 May '17

Ari Seff will present his general exam on Monday, May 8th at 9am in CS 302. The members of his committee are: Han Liu (adviser), Thomas Funkhouser, and Barbara Engelhardt. Everyone is invited to attend his talk, and those faculty wishing to remain for the oral exam following are welcome to do so. His abstract and reading list follow below. Title: Continual Learning in Deep Generative Models Abstract: Developments in deep generative models have allowed for tractable learning of high-dimensional data distributions. Recently introduced frameworks, such as generative adversarial networks and variational autoencoders, can map a noise distribution to the data space, producing realistic samples. In addition, these deep generative models are amenable to conditional training, where a conditional input (e.g., class label, data from another modality) guides the sampling process. At test time, manually selected conditional inputs lead to samples with desired attributes on demand. However, the standard training regime for deep generative models assumes that training data is i.i.d with respect to the conditional inputs. For example, when class labels serve as the conditional inputs, if we wish to extend the capability of a trained network to a new class, training solely on data from the new class will interfere with previously learned distributions. This is due to the general susceptibility of neural networks to catastrophic forgetting when trained on multiple tasks sequentially. Instead, the standard training regime requires that while the network trains on the new class, it simultaneously retrains on the old to circumvent this phenomenon. In this work, we propose a scalable approach to training deep generative models on multiple tasks sequentially, where training data for any previous task (a set of target conditional inputs) is assumed to be inaccessible. To this end, we leverage recent work with discriminative models where plasticity is reduced for network parameters determined to be salient for previously encountered tasks. Experimental evaluation of our approach demonstrates that deep generative models may be extended to settings where the observed data distribution changes over time, enabling continual learning of novel generative tasks. Reading List: Books: I. Goodfellow, Y. Bengio, and A. Courville: Deep learning Papers: Y. LeCun, J. S. Denker, and S. A. Solla: Optimal brain damage I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y. Bengio: Generative adversarial nets D. P. Kingma and M. Welling: Auto-encoding variational bayes J. Kirkpatrick, R. Pascanu, N. C. Rabinowitz, J. Veness, G. Desjardins, A. A. Rusu, K. Milan, J. Quan, T. Ramalho, A. Grabska-Barwinska, D. Hassabis, C. Clopath, D. Kumaran, and R. Hadsell: Overcoming catastrophic forgetting in neural networks M. McCloskey and N. J. Cohen: Catastrophic interference in connectionist networks: The sequential learning problem M. Mirza and S. Osindero: Conditional generative adversarial nets K. Sohn, H. Lee, and X. Yan: Learning structured output representation using deep conditional generative models F. Zenke, B. Poole, and S. Ganguli: Improved multitask learning through synaptic intelligence D. Hoiem and Z. Li: Learning without forgetting A. A. Rusu, N. C. Rabinowitz, G. Desjardins, H. Soyer, J. Kirkpatrick, K. Kavukcuoglu, R. Pascanu, and R. Hadsell: Progressive neural networks

1 0

FPO ANNOUNCEMENT - ALI JAVADIABHARI
by Nicki Gotsis 01 May '17

01 May '17

Department of Electrical Engineering Final Public Oral Examination of Ali JavadiAbhari (advisor: Prof. Margaret Martonosi) Towards a Scalable Software Stack for Resource Estimation and Optimization in General-Purpose Quantum Computers Monday, May 8, 2017 Room J-401 Engineering Quadrangle 10:00 AM All those interested are welcome to attend. A copy of the thesis is on display in the F-wing You can read the thesis abstract here “Please click on the attached .ics file to add this event to your calendar.”

1 0