- talks - lists.cs.princeton.edu

Gregory Gundersen General Exam Presentation Monday, May 21, 2018 at 3:00 pm - CS301
by Barbara A. Mooring 18 May '18

18 May '18

Gregory Gundersen will present his General Exam Presentation on Monday, May 21, 2018 at 3:00 pm in CS301. Committee: Barbara Engelhardt (Adviser) Ryan Adams Mona Singh Title: Multimodal embeddings with structured priors Abstract: Medical images are visually evaluated by experts for key morphological features, but the genes underlying particular features are typically unknown. We propose a multimodal modeling and inference framework that learns modality-specific and interpretable latent representations of both genetic variation and medical image features. The method uses deep canonical correlation analysis (DCCA) to correlate paired embeddings. One modality's embeddings are constrained by a generative prior using a variational autoencoder (VAE). The full model is trained end-to-end with a loss function that combines the DCCA and VAE objectives. We demonstrate utility of this method by exploring the task of jointly learning point cloud clusters of histology slides and gene expression levels and recovering known biological knowledge. Barbara A. Mooring Interim Graduate Coordinator Computer Science Department Princeton University

1 0

Ryan Amos General Exam Presentation May 21, 2018 at 1:00 pm in CS301
by Barbara A. Mooring 17 May '18

17 May '18

Ryan Amos will present General Exam Presentation May 21, 2018 at 1:00 pm in CS301. Title: Verifiable Random Beacons Abstract: Random beacons sources of open, public randomness, and play a critical role in many protocols. Their uses range from proofs of earliest possible creation date to secure multiparty computation to lottery systems. Existing sources of public randomness, including various random beacons, are subject to single points of failure, such that an opportunistic adversary might be able to meaningfully affect the outcome. We develop a random beacon that always requires compromising more than one party to meaningfully affect the outcome. In order to further discourage dishonest behavior, every critical step of our protocol is third-party auditable and backed by economic incentives that make honest behavior more profitable than dishonest behavior. Our method uses a variant of a verifiable random function (VRF) to "scramble" a public input in an unpredictable, thresholded manner, making it difficult for anyone one party to predict how changes to the public input affect the output, or to affect or suppress the output. The proofs of the VRFs are verified by smart contracts to provide monetary reward or penalty, as appropriate, to protocol participants, ensuring honesty is the best fiscal policy. Barbara A. Mooring Interim Graduate Coordinator Computer Science Department Princeton University

1 0

Haoyu Zhang FPO on Thursday, 5/17/2018, at 1:00 pm in CS 302.
by Barbara A. Mooring 17 May '18

17 May '18

Haoyu Zhang will present his FPO on Thursday, 5/17/2018, at 1:00pm in CS 302. The members of his committee are Mike Freedman (advisor), Kyle Jamieson, Wyatt Lloyd (readers), Kai Li, Jennifer Rexford (examiners). Title: Resource Management for Advanced Data Analytics at Large Scale Abstract: The rapidly growing size of data and the complexity of analytics present new challenges for large-scale data analytics systems. Modern distributed computing frameworks need to support not only embarrassingly parallelizable batch jobs, but also advanced applications analyzing text and multimedia data using complex queries and machine learning (ML) models. Given the computation and storage costs of advanced data analytics, resource management is crucial. New applications and workloads expose vastly different characteristics which make traditional scheduling systems inadequate, and at the same time offer great opportunities that lead to new system designs for better performance. In this thesis, we present resource management systems that significantly improve cloud resource efficiency by leveraging the specific characteristics of advanced data analytics applications. We present the design and implementation of the following systems: (i) VideoStorm: a video analytics system that scales to process thousands of vision queries on live video streams over large clusters. VideoStorm's offline profiler generates resource-quality profiles for vision queries, and its online scheduler allocates resources to maximize performance in terms of vision processing quality and lag. (ii) SLAQ: a cluster scheduling system for approximate ML training jobs that aims to maximize the overall model quality. In iterative and exploratory training settings, better models can be obtained faster by directing resources to jobs with the most potential for improvement. SLAQ allocates resources to maximize the cluster-wide quality improvement based on highly-tailored model quality predictions. (iii) Riffle: an optimized shuffle service for big-data analytics frameworks that significantly improves I/O efficiency. The all-to-all data transfer (i.e., shuffle) in modern big-data systems (such as Spark and Hadoop) becomes the scaling bottleneck for multi-stage analytics jobs, due to the superlinear increase in disk I/O operations as data volume grows. Riffle boosts system performance by merging fragmented intermediate files and efficiently scheduling the merge operations. Taken together, this thesis demonstrates a novel set of methods in both job-level and task-level scheduling for building scalable, highly-efficient, and cost-effective resource management systems. We have performed extensive evaluation with real production workloads, and our results show significant improvement in resource efficiency, job completion time, and system throughput for advanced data analytics. Barbara A. Mooring Interim Graduate Coordinator Computer Science Department Princeton University

1 2

Anne Edmundson FPO Wednesday, May 16, 2018 at 2:00 pm CS402
by Barbara A. Mooring 16 May '18

16 May '18

Anne Edmundson will present her FPO on Wednesday, May 16, 2018 at 2:00 pm in CS402. Committee: Nick Feasmster, Adviser - Princeton University Arvind Narayanan - Princeton University Prateek Mittal - Princeton University All are welcomed to attend. Title: Privacy Infrastructure for Content and Communications Abstract: Citizens’ privacy is coming under greater threat as an increasing number of entities can access user data. A powerful adversary, such as a nation-state, can gain access to user data using a broad range of techniques, from privately tapping wires and collecting traffic to serving warrants or subpoenas for user data. Protecting user privacy in the face of these types of activities is challenging. Existing protocol encryption such as TLS is not sufficient, since a wide range of data, from DNS lookups to server access logs, may be visible to eavesdroppers or subject to data requests. In this dissertation, I develop new techniques that demonstrate that three aspects of the existing Internet infrastructure, specifically routing, hosting, and naming, can be used to counter surveillance. First, I study the current state of routing by measuring which countries are on the paths between users and popular websites. I then evaluate di↵erent methods for routing Internet traffic around unfavorable countries, and based on these findings, I design and implement RAN, a lightweight system that routes a client’s web traffic around specified countries with no modifications to client software. Second, I describe modifications to content hosting that prevent a powerful adversary such as a nation-state from gaining access to a user’s requests for certain Web content. In today’s Internet, Content Distribution Networks (CDNs) have rich information both about the content they are serving and the users who are requesting that content. Access to this type of information makes CDNs a target for requests for data about users’ browsing activities. To counter this threat, I developed Oblivious CDN (OCDN), which hides from the CDN both the content it is serving and the users who are requesting that content. In the last part of this dissertation, I explore how the naming infrastructure currently compromises client privacy by looking at conventional DNS as well as onion services. I highlight fundamental issues with both types of domain lookups, and present Oblivious DNS (ODNS) as a new approach to protecting privacy by decoupling client identities from the domains they are looking up. Barbara A. Mooring Interim Graduate Coordinator Computer Science Department Princeton University

1 2

RESENDING: Ben Burgess General Exam Presentation - Friday May 18, 2018 at 10:00 am - CS302
by Barbara A. Mooring 16 May '18

16 May '18

Ben Burgess General Exam Presentation - Friday May 18, 2018 at 10:00 am - CS302 Title: "Security Evaluation of Trunking Abstract; Trunking radio is used as the primary communication system for almost all large commercial and government agencies. These radio system provide more convenience than most cellular phones or conventional two way radios can provide while also in most cases promising the ability to provide completely secure communications. On top of providing the ability to use secure voice, trunking radio vendors offer solutions to rekey the radios over the air in the case that a radio is lost or stolen. In this paper, we select trunking radio solutions from the two dominant vendors in the market, Motorola and Harris. We implement a proof of concept attack that had been previously theorized [Glass 2011] against a real world deployment of the Motorola trunking radio solution. We then extend this attack to compromise the security guarantees the over the air rekey system was thought to provide. We continue on to reverse engineer the protocol specification of the Harris solution which was previously tightly controlled and unavailable for public inspection. We then design an attack against this system which allows for an adversary to both intercept and forge traffic between the subscriber unit and base station. Barbara A. Mooring Interim Graduate Coordinator Computer Science Department Princeton University

1 0

Ben Burgess General Exam Presentation Friday, May 18, 2018 at 10:00 am - CS302
by Barbara A. Mooring 16 May '18

16 May '18

Ben Burgess will present his General Exam Presentation Friday, May 18, 2018 at 10:00 am - CS302 Title: Verifiable Random Beacons Abstract: Random beacons sources of open, public randomness, and play a critical role in many protocols. Their uses range from proofs of earliest possible creation date to secure multiparty computation to lottery systems. Existing sources of public randomness, including various random beacons, are subject to single points of failure, such that an opportunistic adversary might be able to meaningfully affect the outcome. We develop a random beacon that always requires compromising more than one party to meaningfully affect the outcome. In order to further discourage dishonest behavior, every critical step of our protocol is third-party auditable and backed by economic incentives that make honest behavior more profitable than dishonest behavior. Our method uses a variant of a verifiable random function (VRF) to "scramble" a public input in an unpredictable, thresholded manner, making it difficult for anyone one party to predict how changes to the public input affect the output, or to affect or suppress the output. The proofs of the VRFs are verified by smart contracts to provide monetary reward or penalty, as appropriate, to protocol participants, ensuring honesty is the best fiscal policy. Barbara A. Mooring Interim Graduate Coordinator Computer Science Department Princeton University

1 0

Jude Nelson FPO on Tuesday, May 22 at 2:00 pm in CS402
by Barbara A. Mooring 15 May '18

15 May '18

Jude Nelson will present his FPO on Tuesday, May 22 at 2:00 pm in CS 402. Committee: Larry Peterson, Adviser Princeton University Kai Li, Princeton University Brian Kernighan, Princeton University John H. Hartman, University of Arizona All are welcomed to attend. Title: Wide-area Software-defined Storage Abstract: The proliferation of commodity cloud services helps developers build wide-area “system-of-systems” applications by harnessing cloud storage, CDNs, and public datasets as reusable building blocks. But to do so, developers must contend with two long-term challenges. First, whenever developers change storage providers, they must work to preserve the application’s expected storage semantics, i.e. the rules governing how the application expects the storage provider to handle its reads and writes. Today, changing storage providers is costly, because developers need to patch the application to make it compatible with the new provider’s data consistency model, access controls, replica placement strategies, and so on. At the same time, users have certain expectations about how their data will be used, which the application must meet. For example, depending on the application, users may expect that their data will be kept private from other users, that their data will be exportable to other applications, that accesses to their data will be logged in an auditable way, and so on. In the limit, each user’s expectations represent an implicit policy constraining how their data can be stored. Honoring these policies is difficult for developers who rely on third-party storage providers because the storage provider is often unaware of them. This thesis addresses these challenges with a wide-area storage protocol, called “software-defined storage” (SDS), that runs in-between applications and cloud ser- vices. SDS-enabled applications do not host data, but instead let users bring their preferred cloud services to the application. By taking a user-centric approach to hosting data, users are empowered to programmatically specify their policies inde- pendent of their applications and select services that will honor them. To support this approach and to tolerate service provider changes, SDS empowers developers to programmatically specify their application’s storage semantics independent of storage providers. This thesis presents the design principles for SDS, and validates their real-world applicability with two SDS implementations and several non-trivial applications built on top of them. Most of these applications are used in production today. This thesis presents early performance results of its SDS implementations and uses real-world experiences to show how to make the most of SDS. Barbara A. Mooring Interim Graduate Coordinator Computer Science Department Princeton University

1 1

REMINDER - Arunesh Mathur General Exam Presentation TODAY, May 15, 2018 at 2:00 pm, CS402
by Barbara A. Mooring 15 May '18

15 May '18

REMINDER - Arunesh Mathur General Exam Presentation TODAY, May 15, 2018 at 2:00 pm, CS402 Title: Examining Advertising Dark Patterns on Social Media: An Empirical Study of Affiliate Marketing Endorsements Abstract: Dark patterns are persuasive techniques in interface design that nudge users towards making sub-optimal decisions in order to benefit the creator of the interface. A common dark pattern that manifests in online platforms involves disguising advertisements as testimonials, endorsements, or reviews to drive sales and clicks. Because users are often ill-equipped to identify such advertisements, the Federal Trade Commission (FTC) in the United States requires that these advertisements be adequately disclosed to avoid user deception and harm. In this study, we examined disclosures accompanying affiliate marketing, a type of endorsement marketing used by content creators on social media platforms. Specifically, we asked how many content creators disclose their affiliate marketing content to users, whether content creators follow the FTC’s disclosure guidelines in their disclosures, and whether users comprehend the disclosure statements. By examining over half a million YouTube videos and two million Pinterest pins, we found that only roughly one-tenth of affiliate marketing content on both platforms contains any form of disclosures. In the following user study, we discovered that users fail to understand shorter, non-explanatory disclosures. Based on our findings, we make various design and policy suggestions that can improve disclosure practices. Barbara A. Mooring Interim Graduate Coordinator Computer Science Department Princeton University

1 0

Andrew Or General Exam Presentation Wednesday, May 16, 2018 at 10:00 am - CS302
by Barbara A. Mooring 14 May '18

14 May '18

Andrew Or General Exam Presentation Wednesday, May 16, 2018 at 10:00 am - CS302 Title: Quality-aware cluster scheduling Abstract: Modern data analytics is increasingly concerned with quick response times. With the recent growth of data far exceeding the rate at which hardware has advanced, many applications have begun to explore options that return intermediate results early. For example, approximate query processing systems begin returning inexact answers before having processed the entire input data, and machine learning practitioners often stop the training process early once they are sufficiently satisfied with the training error. However, existing cluster schedulers are not well equipped for these kinds of applications. In particular, state-of-the-art scheduling solutions are primarily focused on resource fairness when making decisions, treating the applications running on the cluster as black boxes. This approach, though widely popular, often wastes a significant amount of resources on applications that have been running for a long time and are no longer making much progress due to diminishing marginal utility. This talk presents an alternative scheduling approach that makes decisions based on application utility. The intuition behind this approach is that prioritizing fresh applications can lead to cluster-wide utility improvement. We have implemented such a scheduler in Spark and will discuss its effectiveness in the context of approximate query processing and machine learning. Barbara A. Mooring Interim Graduate Coordinator Computer Science Department Princeton University

1 0

Arnesh Mathur General Exam Presentation Tuesday, May 15, 2018 at 2:00 pm - CS402
by Barbara A. Mooring 14 May '18

14 May '18

Arnesh Mathur General Exam Presentation Tuesday, May 15, 2018 at 2:00 pm - CS402 Title: Examining Advertising Dark Patterns on Social Media: An Empirical Study of Affiliate Marketing Endorsements Abstract: Dark patterns are persuasive techniques in interface design that nudge users towards making sub-optimal decisions in order to benefit the creator of the interface. A common dark pattern that manifests in online platforms involves disguising advertisements as testimonials, endorsements, or reviews to drive sales and clicks. Because users are often ill-equipped to identify such advertisements, the Federal Trade Commission (FTC) in the United States requires that these advertisements be adequately disclosed to avoid user deception and harm. In this study, we examined disclosures accompanying affiliate marketing, a type of endorsement marketing used by content creators on social media platforms. Specifically, we asked how many content creators disclose their affiliate marketing content to users, whether content creators follow the FTC’s disclosure guidelines in their disclosures, and whether users comprehend the disclosure statements. By examining over half a million YouTube videos and two million Pinterest pins, we found that only roughly one-tenth of affiliate marketing content on both platforms contains any form of disclosures. In the following user study, we discovered that users fail to understand shorter, non-explanatory disclosures. Based on our findings, we make various design and policy suggestions that can improve disclosure practices. Barbara A. Mooring Interim Graduate Coordinator Computer Science Department Princeton University

1 0