Proceedings of the 2020 ACM-IMS on Foundations of Data Science Conference最新文献_第2页

StyleCAPTCHA

Proceedings of the 2020 ACM-IMS on Foundations of Data Science Conference Pub Date : 2020-10-18 DOI: 10.1145/3412815.3416895

Haitian Chen, Bai Jiang, Hao Chen

{"title":"StyleCAPTCHA","authors":"Haitian Chen, Bai Jiang, Hao Chen","doi":"10.1145/3412815.3416895","DOIUrl":"https://doi.org/10.1145/3412815.3416895","url":null,"abstract":"CAPTCHAs are widely deployed for bot detection. Many CAPTCHAs are based on visual perception tasks such as text and objection classification. However, they are under serious threat from advanced visual perception technologies based on deep convolutional networks (DCNs). We propose a novel CAPTCHA, called StyleCAPTCHA, that asks a user to classify stylized human versus animal face images. StyleCAPTCHA creates each stylized image by combining the content representations of a human or animal face image and the style representations of a reference image. Both the original face image and the style reference image are hidden from the user. To defend against attacks using DCNs, the StyleCAPTCHA service changes the style regularly. To adapt to the new styles, the attacker has to repeatedly train or retrain her DCNs, but since the attacker has insufficient training examples, she cannot train her DCNs well. We also propose Classifier Cross-task Transferability to measure the transferability of a classifier from its original task to another task. This metric allows us to arrange the schedule of styles and to limit the transferability of attackers' DCNs across classification tasks using different styles. Our evaluation shows that StyleCAPTCHA defends against state-of-the-art face detectors and against general DCN classifiers effectively.","PeriodicalId":176130,"journal":{"name":"Proceedings of the 2020 ACM-IMS on Foundations of Data Science Conference","volume":"79 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-10-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116096126","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 1

Session details: Keynote Talk II 会议详情:主题演讲II

Proceedings of the 2020 ACM-IMS on Foundations of Data Science Conference Pub Date : 2020-10-18 DOI: 10.1145/3429734

Jeannette M. Wing

引用次数: 0

Toward Communication Efficient Adaptive Gradient Method 通信高效自适应梯度方法研究

Proceedings of the 2020 ACM-IMS on Foundations of Data Science Conference Pub Date : 2020-10-18 DOI: 10.1145/3412815.3416891

Xiangyi Chen, Xiaoyun Li, P. Li

引用次数: 32

Applying Algorithmic Accountability Frameworks with Domain-specific Codes of Ethics: A Case Study in Ecosystem Forecasting for Shellfish Toxicity in the Gulf of Maine 应用特定领域道德规范的算法问责框架:缅因州湾贝类毒性生态系统预测的案例研究

Proceedings of the 2020 ACM-IMS on Foundations of Data Science Conference Pub Date : 2020-10-18 DOI: 10.1145/3412815.3416897

Isabella Grasso, David Russell, Abigail V. Matthews, Jeanna Neefe Matthews

{"title":"Applying Algorithmic Accountability Frameworks with Domain-specific Codes of Ethics: A Case Study in Ecosystem Forecasting for Shellfish Toxicity in the Gulf of Maine","authors":"Isabella Grasso, David Russell, Abigail V. Matthews, Jeanna Neefe Matthews","doi":"10.1145/3412815.3416897","DOIUrl":"https://doi.org/10.1145/3412815.3416897","url":null,"abstract":"Ecological forecasts are used to inform decisions that can havesignificant impacts on the lives of individuals and on the healthof ecosystems. These forecasts, or models, embody the ethics oftheir creators as well as many seemingly arbitrary implementationchoices made along the way. They can contain implementationerrors as well as reflect patterns of bias learned when ingestingdatasets derived from past biased decision making. Principles andframeworks for algorithmic accountability allow a wide range ofstakeholders to place the results of models and software systemsinto context. We demonstrate how the combination of algorithmicaccountability frameworks and domain-specific codes of ethics helpanswer calls to uphold fairness and human values, specifically indomains that utilize machine learning algorithms. This helps avoidmany of the unintended consequences that can result from deploy-ing \"black box\" systems to solve complex problems. In this paper,we discuss our experience applying algorithmic accountability prin-ciples and frameworks to ecosystem forecasting, focusing on a casestudy forecasting shellfish toxicity in the Gulf of Maine. We adaptexisting frameworks such as Datasheets for Datasets and ModelCards for Model Reporting from their original focus on personallyidentifiable private data to include public datasets, such as thoseoften used in ecosystem forecasting applications, to audit the casestudy. We show how high level algorithmic accountability frame-works and domain level codes of ethics compliment each other,incentivizing more transparency, accountability, and fairness inautomated decision-making systems.","PeriodicalId":176130,"journal":{"name":"Proceedings of the 2020 ACM-IMS on Foundations of Data Science Conference","volume":"77 1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-10-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126120695","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 4

ADAGES

Proceedings of the 2020 ACM-IMS on Foundations of Data Science Conference Pub Date : 2020-10-18 DOI: 10.1145/3412815.3416881

Yu Gui

引用次数: 11

Large Very Dense Subgraphs in a Stream of Edges 边流中的大型非常密集子图

Proceedings of the 2020 ACM-IMS on Foundations of Data Science Conference Pub Date : 2020-10-15 DOI: 10.1145/3412815.3416884

Claire Mathieu, Michel de Rougemont

{"title":"Large Very Dense Subgraphs in a Stream of Edges","authors":"Claire Mathieu, Michel de Rougemont","doi":"10.1145/3412815.3416884","DOIUrl":"https://doi.org/10.1145/3412815.3416884","url":null,"abstract":"We study the detection and the reconstruction of a large very dense subgraph in a social graph with n nodes and m edges given as a stream of edges, when the graph follows a power law degree distribution, in the regime when $m=O(n. łog n)$. A subgraph is very dense if its edge density is comparable to a clique. We uniformly sample the edges with a Reservoir of size $k=O(sqrtn.łog n)$. The detection algorithm of a large very dense subgraph checks whether the Reservoir has a giant component. We show that if the graph contains a very dense subgraph of size $Ømega(sqrtn )$, then the detection algorithm is almost surely correct. On the other hand, a random graph that follows a power law degree distribution almost surely has no large very dense subgraph, and the detection algorithm is almost surely correct. We define a new model of random graphs which follow a power law degree distribution and have large very dense subgraphs. We then show that on this class of random graphs we can reconstruct a good approximation of the very dense subgraph with high probability. We generalize these results to dynamic graphs defined by sliding windows in a stream of edges.","PeriodicalId":176130,"journal":{"name":"Proceedings of the 2020 ACM-IMS on Foundations of Data Science Conference","volume":"29 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-10-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114771178","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 4

Congenial Differential Privacy under Mandated Disclosure 强制披露下的同类差异隐私

Proceedings of the 2020 ACM-IMS on Foundations of Data Science Conference Pub Date : 2020-08-24 DOI: 10.1145/3412815.3416892

Ruobin Gong, X. Meng

{"title":"Congenial Differential Privacy under Mandated Disclosure","authors":"Ruobin Gong, X. Meng","doi":"10.1145/3412815.3416892","DOIUrl":"https://doi.org/10.1145/3412815.3416892","url":null,"abstract":"Differentially private data releases are often required to satisfy a set of external constraints that reflect the legal, ethical, and logical mandates to which the data curator is obligated. The enforcement of constraints, when treated as post-processing, adds an extra phase in the production of privatized data. It is well understood in the theory of multi-phase processing that congeniality, a form of procedural compatibility between phases, is a prerequisite for the end users to straightforwardly obtain statistically valid results. Congenial differential privacy is theoretically principled, which facilitates transparency and intelligibility of the mechanism that would otherwise be undermined by ad-hoc post-processing procedures. We advocate for the systematic integration of mandated disclosure into the design of the privacy mechanism via standard probabilistic conditioning on the invariant margins. Conditioning automatically renders congeniality because any extra post-processing phase becomes unnecessary. We provide both initial theoretical guarantees and a Markov chain algorithm for our proposal. We also discuss intriguing theoretical issues that arise in comparing congenital differential privacy and optimization-based post-processing, as well as directions for further research.","PeriodicalId":176130,"journal":{"name":"Proceedings of the 2020 ACM-IMS on Foundations of Data Science Conference","volume":"58 10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126078246","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 22

Transforming Probabilistic Programs for Model Checking 用于模型检验的转换概率程序

Proceedings of the 2020 ACM-IMS on Foundations of Data Science Conference Pub Date : 2020-08-21 DOI: 10.1145/3412815.3416896

Ryan Bernstein, Matthijs V'ak'ar, Jeannette M. Wing

{"title":"Transforming Probabilistic Programs for Model Checking","authors":"Ryan Bernstein, Matthijs V'ak'ar, Jeannette M. Wing","doi":"10.1145/3412815.3416896","DOIUrl":"https://doi.org/10.1145/3412815.3416896","url":null,"abstract":"Probabilistic programming is perfectly suited to reliable and transparent data science, as it allows the user to specify their models in a high-level language without worrying about the complexities of how to fit the models. Static analysis of probabilistic programs presents even further opportunities for enabling a high-level style of programming, by automating time-consuming and error-prone tasks. We apply static analysis to probabilistic programs to automate large parts of two crucial model checking methods: Prior Predictive Checks and Simulation-Based Calibration. Our method transforms a probabilistic program specifying a density function into an efficient forward-sampling form. To achieve this transformation, we extract a factor graph from a probabilistic program using static analysis, generate a set of proposal directed acyclic graphs using a SAT solver, select a graph which will produce provably correct sampling code, then generate one or more sampling programs. We allow minimal user interaction to broaden the scope of application beyond what is possible with static analysis alone. We present an implementation targeting the popular Stan probabilistic programming language, automating large parts of a robust Bayesian workflow for a wide community of probabilistic programming users.","PeriodicalId":176130,"journal":{"name":"Proceedings of the 2020 ACM-IMS on Foundations of Data Science Conference","volume":"36 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114682345","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Non-Uniform Sampling of Fixed Margin Binary Matrices 固定边界二值矩阵的非均匀抽样

Proceedings of the 2020 ACM-IMS on Foundations of Data Science Conference Pub Date : 2020-07-29 DOI: 10.1145/3412815.3416887

A. Fout, B. Fosdick, Matthew P. Hitt

引用次数: 1

On Reinforcement Learning for Turn-based Zero-sum Markov Games 基于回合制零和马尔可夫博弈的强化学习研究

Proceedings of the 2020 ACM-IMS on Foundations of Data Science Conference Pub Date : 2020-02-25 DOI: 10.1145/3412815.3416888

D. Shah, Varun Somani, Qiaomin Xie, Zhi Xu

引用次数: 8