Shuqing Bian, Wayne Xin Zhao, Kun Zhou, Xu Chen, Jing Cai, Yancheng He, Xingji Luo, Ji-rong Wen
The evolution of mobile apps has greatly changed the way that we live. It becomes increasingly important to understand and model the users on mobile apps. Instead of focusing on some specific app alone, it has become a popular paradigm to study the user behavior on various mobile apps in a symbiotic environment. In this paper, we study the task of user representation learning with both macro and micro interaction data on mobile apps. Specifically, macro and micro interaction refer to user-app interaction or user-item interaction on some specific app, respectively. By combining the two kinds of user data, it is expected to derive a more comprehensive, robust user representation model on mobile apps. In order to effectively fuse the information across the macro and micro views, we propose a novel macro-micro fusion network for user representation learning on mobile apps. With a Transformer architecture as the base model, we design a representation fusion component that is able to capture the category-based semantic alignment at the user level. After such semantic alignment, the information across the two views can be adaptively fused in our approach. Furthermore, we adopt mutual information maximization to derive a self-supervised loss to enhance the learning of our fusion network. Extensive experiments with three downstream tasks on two real-world datasets have demonstrated the effectiveness of our approach.
{"title":"A Novel Macro-Micro Fusion Network for User Representation Learning on Mobile Apps","authors":"Shuqing Bian, Wayne Xin Zhao, Kun Zhou, Xu Chen, Jing Cai, Yancheng He, Xingji Luo, Ji-rong Wen","doi":"10.1145/3442381.3450109","DOIUrl":"https://doi.org/10.1145/3442381.3450109","url":null,"abstract":"The evolution of mobile apps has greatly changed the way that we live. It becomes increasingly important to understand and model the users on mobile apps. Instead of focusing on some specific app alone, it has become a popular paradigm to study the user behavior on various mobile apps in a symbiotic environment. In this paper, we study the task of user representation learning with both macro and micro interaction data on mobile apps. Specifically, macro and micro interaction refer to user-app interaction or user-item interaction on some specific app, respectively. By combining the two kinds of user data, it is expected to derive a more comprehensive, robust user representation model on mobile apps. In order to effectively fuse the information across the macro and micro views, we propose a novel macro-micro fusion network for user representation learning on mobile apps. With a Transformer architecture as the base model, we design a representation fusion component that is able to capture the category-based semantic alignment at the user level. After such semantic alignment, the information across the two views can be adaptively fused in our approach. Furthermore, we adopt mutual information maximization to derive a self-supervised loss to enhance the learning of our fusion network. Extensive experiments with three downstream tasks on two real-world datasets have demonstrated the effectiveness of our approach.","PeriodicalId":106672,"journal":{"name":"Proceedings of the Web Conference 2021","volume":"106 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-04-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117193466","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}
Single Sign-On (SSO) has been widely adopted for online authentication due to its favorable usability and security. However, it also introduces a single point of failure since all service providers fully trust the identity of a user created by the SSO identity provider. In this paper, we investigate the identity-account inconsistency threat, a new SSO vulnerability that can cause the compromise of online accounts. The vulnerability exists because current SSO systems highly rely on a user’s email address to bind an account with a real identity, but ignore the fact that email addresses might be reused by other users. We reveal that under the SSO authentication, such inconsistency allows an adversary controlling a reused email address to take over associated online accounts without knowing any credentials like passwords. Specifically, we first conduct a measurement study on the account management policies for multiple cloud email providers, showing the feasibility of acquiring previously used email accounts. We further perform a systematic study on 100 popular websites using the Google business email service with our own domain address and demonstrate that most online accounts can be compromised by exploiting this inconsistency vulnerability. To shed light on email reuse in the wild, we analyze the commonly used naming conventions that lead to a wide existence of potential email address collisions, and conduct a case study on the account policies of U.S. universities. Finally, we propose several useful practices for end-users, service providers, and identity providers to protect against this identity-account inconsistency threat.
{"title":"An Investigation of Identity-Account Inconsistency in Single Sign-On","authors":"Guannan Liu, Xing Gao, Haining Wang","doi":"10.1145/3442381.3450085","DOIUrl":"https://doi.org/10.1145/3442381.3450085","url":null,"abstract":"Single Sign-On (SSO) has been widely adopted for online authentication due to its favorable usability and security. However, it also introduces a single point of failure since all service providers fully trust the identity of a user created by the SSO identity provider. In this paper, we investigate the identity-account inconsistency threat, a new SSO vulnerability that can cause the compromise of online accounts. The vulnerability exists because current SSO systems highly rely on a user’s email address to bind an account with a real identity, but ignore the fact that email addresses might be reused by other users. We reveal that under the SSO authentication, such inconsistency allows an adversary controlling a reused email address to take over associated online accounts without knowing any credentials like passwords. Specifically, we first conduct a measurement study on the account management policies for multiple cloud email providers, showing the feasibility of acquiring previously used email accounts. We further perform a systematic study on 100 popular websites using the Google business email service with our own domain address and demonstrate that most online accounts can be compromised by exploiting this inconsistency vulnerability. To shed light on email reuse in the wild, we analyze the commonly used naming conventions that lead to a wide existence of potential email address collisions, and conduct a case study on the account policies of U.S. universities. Finally, we propose several useful practices for end-users, service providers, and identity providers to protect against this identity-account inconsistency threat.","PeriodicalId":106672,"journal":{"name":"Proceedings of the Web Conference 2021","volume":"66 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-04-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117353666","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}
Test collection has been a crucial factor for developing information retrieval systems. Constructing a test collection requires annotators to assess the relevance of massive query-document pairs. Relevance annotations acquired through crowdsourcing platforms alleviate the enormous cost of this process but they are often noisy. Existing models to denoise crowd annotations mostly assume that annotations are generated independently, based on which a probabilistic graphical model is designed to model the annotation generation process. However, tasks are often correlated with each other in reality. It is an understudied problem whether and how task correlation helps in denoising crowd annotations. In this paper, we relax the independence assumption to model task correlation in terms of relevance. We propose a new crowd annotation generation model named CrowdGP, where true relevance labels, annotator competence, annotator’s bias towards relevancy, task difficulty, and task’s bias towards relevancy are modelled through a Gaussian process and multiple Gaussian variables respectively. The CrowdGP model shows better performance in terms of interring true relevance labels compared with state-of-the-art baselines on two crowdsourcing datasets on relevance. The experiments also demonstrate its effectiveness in terms of selecting new tasks for future crowd annotation, which is a new functionality of CrowdGP. Ablation studies indicate that the effectiveness is attributed to the modelling of task correlation based on the auxiliary information of tasks and the prior relevance information of documents to queries.
{"title":"CrowdGP: a Gaussian Process Model for Inferring Relevance from Crowd Annotations","authors":"Dan Li, Zhaochun Ren, E. Kanoulas","doi":"10.1145/3442381.3450047","DOIUrl":"https://doi.org/10.1145/3442381.3450047","url":null,"abstract":"Test collection has been a crucial factor for developing information retrieval systems. Constructing a test collection requires annotators to assess the relevance of massive query-document pairs. Relevance annotations acquired through crowdsourcing platforms alleviate the enormous cost of this process but they are often noisy. Existing models to denoise crowd annotations mostly assume that annotations are generated independently, based on which a probabilistic graphical model is designed to model the annotation generation process. However, tasks are often correlated with each other in reality. It is an understudied problem whether and how task correlation helps in denoising crowd annotations. In this paper, we relax the independence assumption to model task correlation in terms of relevance. We propose a new crowd annotation generation model named CrowdGP, where true relevance labels, annotator competence, annotator’s bias towards relevancy, task difficulty, and task’s bias towards relevancy are modelled through a Gaussian process and multiple Gaussian variables respectively. The CrowdGP model shows better performance in terms of interring true relevance labels compared with state-of-the-art baselines on two crowdsourcing datasets on relevance. The experiments also demonstrate its effectiveness in terms of selecting new tasks for future crowd annotation, which is a new functionality of CrowdGP. Ablation studies indicate that the effectiveness is attributed to the modelling of task correlation based on the auxiliary information of tasks and the prior relevance information of documents to queries.","PeriodicalId":106672,"journal":{"name":"Proceedings of the Web Conference 2021","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-04-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127267322","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}
In this paper, we design, implement, and evaluate , which is to our knowledge the first consumer-class system that streams panoramic videos far beyond the ultra high-definition resolution (up to 16K) to mobile devices, offering truly immersive experiences. Such an immense resolution makes streaming video-on-demand (VoD) content extremely resource-demanding. To tackle this challenge, introduces a novel framework that leverages an edge server to perform efficient, intelligent, and quality-guaranteed content transcoding, by extracting from panoramic frames the viewport stream that will be delivered to the client. To support real-time transcoding of 16K content, employs several key mechanisms such as dual-GPU acceleration, lossless viewport extraction, deep viewport prediction, and a two-layer streaming design. Our extensive evaluations using real users’ viewport movement data indicate that outperforms existing solutions, and can smoothly stream 16K panoramic videos to mobile devices over diverse wireless networks including WiFi, LTE, and mmWave 5G.
{"title":"DeepVista: 16K Panoramic Cinema on Your Mobile Device","authors":"Wenxiao Zhang, Feng Qian, B. Han, P. Hui","doi":"10.1145/3442381.3449829","DOIUrl":"https://doi.org/10.1145/3442381.3449829","url":null,"abstract":"In this paper, we design, implement, and evaluate , which is to our knowledge the first consumer-class system that streams panoramic videos far beyond the ultra high-definition resolution (up to 16K) to mobile devices, offering truly immersive experiences. Such an immense resolution makes streaming video-on-demand (VoD) content extremely resource-demanding. To tackle this challenge, introduces a novel framework that leverages an edge server to perform efficient, intelligent, and quality-guaranteed content transcoding, by extracting from panoramic frames the viewport stream that will be delivered to the client. To support real-time transcoding of 16K content, employs several key mechanisms such as dual-GPU acceleration, lossless viewport extraction, deep viewport prediction, and a two-layer streaming design. Our extensive evaluations using real users’ viewport movement data indicate that outperforms existing solutions, and can smoothly stream 16K panoramic videos to mobile devices over diverse wireless networks including WiFi, LTE, and mmWave 5G.","PeriodicalId":106672,"journal":{"name":"Proceedings of the Web Conference 2021","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-04-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126153943","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}
Feiyang Pan, Haoming Li, Xiang Ao, Wei Wang, Yanrong Kang, Ao Tan, Qing He
The exploration/exploitation (E&E) dilemma lies at the core of interactive systems such as online advertising, for which contextual bandit algorithms have been proposed. Bayesian approaches provide guided exploration via uncertainty estimation, but the applicability is often limited due to over-simplified assumptions. Non-Bayesian bootstrap methods, on the other hand, can apply to complex problems by using deep reward models, but lack a clear guidance to the exploration behavior. It still remains largely unsolved to develop a practical method for complex deep contextual bandits. In this paper, we introduce Guided Bootstrap (GuideBoot), combining the best of both worlds. GuideBoot provides explicit guidance to the exploration behavior by training multiple models over both real samples and noisy samples with fake labels, where the noise is added according to the predictive uncertainty. The proposed method is efficient as it can make decisions on-the-fly by utilizing only one randomly chosen model, but is also effective as we show that it can be viewed as a non-Bayesian approximation of Thompson sampling. Moreover, we extend it to an online version that can learn solely from streaming data, which is favored in real applications. Extensive experiments on both synthetic tasks and large-scale advertising environments show that GuideBoot achieves significant improvements against previous state-of-the-art methods.
{"title":"GuideBoot: Guided Bootstrap for Deep Contextual Banditsin Online Advertising","authors":"Feiyang Pan, Haoming Li, Xiang Ao, Wei Wang, Yanrong Kang, Ao Tan, Qing He","doi":"10.1145/3442381.3449987","DOIUrl":"https://doi.org/10.1145/3442381.3449987","url":null,"abstract":"The exploration/exploitation (E&E) dilemma lies at the core of interactive systems such as online advertising, for which contextual bandit algorithms have been proposed. Bayesian approaches provide guided exploration via uncertainty estimation, but the applicability is often limited due to over-simplified assumptions. Non-Bayesian bootstrap methods, on the other hand, can apply to complex problems by using deep reward models, but lack a clear guidance to the exploration behavior. It still remains largely unsolved to develop a practical method for complex deep contextual bandits. In this paper, we introduce Guided Bootstrap (GuideBoot), combining the best of both worlds. GuideBoot provides explicit guidance to the exploration behavior by training multiple models over both real samples and noisy samples with fake labels, where the noise is added according to the predictive uncertainty. The proposed method is efficient as it can make decisions on-the-fly by utilizing only one randomly chosen model, but is also effective as we show that it can be viewed as a non-Bayesian approximation of Thompson sampling. Moreover, we extend it to an online version that can learn solely from streaming data, which is favored in real applications. Extensive experiments on both synthetic tasks and large-scale advertising environments show that GuideBoot achieves significant improvements against previous state-of-the-art methods.","PeriodicalId":106672,"journal":{"name":"Proceedings of the Web Conference 2021","volume":"123 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-04-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115555573","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}
Lihong Wang, Juwei Yue, Shu Guo, Jiawei Sheng, Qianren Mao, Zhenyu Chen, Shenghai Zhong, Chen Li
Script event prediction (SEP) aims to choose a correct subsequent event from a candidate list, given a chain of ordered context events. Event representation learning has been proposed and successfully applied to this task. Most previous methods learning representations mainly focus on coarse-grained connections at event or chain level, while ignoring more fine-grained connections between events. Here we propose a novel framework which can enhance the representation learning of events by mining their connections at multiple granularity levels, including argument level, event level and chain level. In our method, we first employ a masked self-attention mechanism to model the relations between the components of events (i.e. arguments). Then, a directed graph convolutional network is further utilized to model the temporal or causal relations between events in the chain. Finally, we introduce an attention module to the context event chain, so as to dynamically aggregate context events with respect to the current candidate event. By fusing threefold connections in a unified framework, our approach can learn more accurate argument/event/chain representations, and thus leads to better prediction performance. Comprehensive experiment results on public New York Times corpus demonstrate that our model outperforms other state-of-the-art baselines. Our code is available in https://github.com/YueAWu/MCer.
{"title":"Multi-level Connection Enhanced Representation Learning for Script Event Prediction","authors":"Lihong Wang, Juwei Yue, Shu Guo, Jiawei Sheng, Qianren Mao, Zhenyu Chen, Shenghai Zhong, Chen Li","doi":"10.1145/3442381.3449894","DOIUrl":"https://doi.org/10.1145/3442381.3449894","url":null,"abstract":"Script event prediction (SEP) aims to choose a correct subsequent event from a candidate list, given a chain of ordered context events. Event representation learning has been proposed and successfully applied to this task. Most previous methods learning representations mainly focus on coarse-grained connections at event or chain level, while ignoring more fine-grained connections between events. Here we propose a novel framework which can enhance the representation learning of events by mining their connections at multiple granularity levels, including argument level, event level and chain level. In our method, we first employ a masked self-attention mechanism to model the relations between the components of events (i.e. arguments). Then, a directed graph convolutional network is further utilized to model the temporal or causal relations between events in the chain. Finally, we introduce an attention module to the context event chain, so as to dynamically aggregate context events with respect to the current candidate event. By fusing threefold connections in a unified framework, our approach can learn more accurate argument/event/chain representations, and thus leads to better prediction performance. Comprehensive experiment results on public New York Times corpus demonstrate that our model outperforms other state-of-the-art baselines. Our code is available in https://github.com/YueAWu/MCer.","PeriodicalId":106672,"journal":{"name":"Proceedings of the Web Conference 2021","volume":"151 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-04-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116094873","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}
Motivated by the goal of designing interventions for softening polarized opinions on the Web, and building on results from psychology, we hypothesized that people would be moved more easily towards opposing opinions when the latter were voiced by a celebrity they like, rather than by a celebrity they dislike. We tested this hypothesis in a survey-based randomized controlled trial in which we exposed respondents to opinions that were randomly assigned to one of four spokespersons each: a disagreeing but liked celebrity, a disagreeing and disliked celebrity, a disagreeing expert, and an agreeing but disliked celebrity. After the treatment, we measured changes in the respondents’ opinions, empathy towards the spokespersons, and use of affective language. Unlike hypothesized, no softening of opinions was observed regardless of the respondents’ attitudes towards the celebrity. Instead, we found strong evidence of a hardening of pretreatment opinions when a disagreeing opinion was attributed to an expert or when an agreeing opinion was attributed to a disliked celebrity. We also observed a pronounced reduction in empathy for disagreeing spokespersons, indicating a punitive response. The only celebrity for whom, on average, empathy remained unchanged was the one who agreed, even though they were disliked. Our results could be explained as a reaction to violated expectations towards experts and as a perceived breach of trust by liked celebrities. They confirm that naïve strategies at mediation may not yield intended results, and how difficult it is to depolarize—and how easy it is to further polarize or provoke emotional responses.
{"title":"Interventions for Softening Can Lead to Hardening of Opinions: Evidence from a Randomized Controlled Trial","authors":"A. Spitz, A. Abu-Akel, R. West","doi":"10.1145/3442381.3450019","DOIUrl":"https://doi.org/10.1145/3442381.3450019","url":null,"abstract":"Motivated by the goal of designing interventions for softening polarized opinions on the Web, and building on results from psychology, we hypothesized that people would be moved more easily towards opposing opinions when the latter were voiced by a celebrity they like, rather than by a celebrity they dislike. We tested this hypothesis in a survey-based randomized controlled trial in which we exposed respondents to opinions that were randomly assigned to one of four spokespersons each: a disagreeing but liked celebrity, a disagreeing and disliked celebrity, a disagreeing expert, and an agreeing but disliked celebrity. After the treatment, we measured changes in the respondents’ opinions, empathy towards the spokespersons, and use of affective language. Unlike hypothesized, no softening of opinions was observed regardless of the respondents’ attitudes towards the celebrity. Instead, we found strong evidence of a hardening of pretreatment opinions when a disagreeing opinion was attributed to an expert or when an agreeing opinion was attributed to a disliked celebrity. We also observed a pronounced reduction in empathy for disagreeing spokespersons, indicating a punitive response. The only celebrity for whom, on average, empathy remained unchanged was the one who agreed, even though they were disliked. Our results could be explained as a reaction to violated expectations towards experts and as a perceived breach of trust by liked celebrities. They confirm that naïve strategies at mediation may not yield intended results, and how difficult it is to depolarize—and how easy it is to further polarize or provoke emotional responses.","PeriodicalId":106672,"journal":{"name":"Proceedings of the Web Conference 2021","volume":"57 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-04-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116133686","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}
Multivariate Hawkes Processes (MHPs) are an important class of temporal point processes that have enabled key advances in understanding and predicting social information systems. However, due to their complex modeling of temporal dependencies, MHPs have proven to be notoriously difficult to scale, what has limited their applications to relatively small domains. In this work, we propose a novel model and computational approach to overcome this important limitation. By exploiting a characteristic sparsity pattern in real-world diffusion processes, we show that our approach allows to compute the exact likelihood and gradients of an MHP – independently of the ambient dimensions of the underlying network. We show on synthetic and real-world datasets that our method does not only achieve state-of-the-art modeling results, but also improves runtime performance by multiple orders of magnitude on sparse event sequences. In combination with easily interpretable latent variables and influence structures, this allows us to analyze diffusion processes in networks at previously unattainable scale.
{"title":"Modeling Sparse Information Diffusion at Scale via Lazy Multivariate Hawkes Processes","authors":"Maximilian Nickel, Matt Le","doi":"10.1145/3442381.3450094","DOIUrl":"https://doi.org/10.1145/3442381.3450094","url":null,"abstract":"Multivariate Hawkes Processes (MHPs) are an important class of temporal point processes that have enabled key advances in understanding and predicting social information systems. However, due to their complex modeling of temporal dependencies, MHPs have proven to be notoriously difficult to scale, what has limited their applications to relatively small domains. In this work, we propose a novel model and computational approach to overcome this important limitation. By exploiting a characteristic sparsity pattern in real-world diffusion processes, we show that our approach allows to compute the exact likelihood and gradients of an MHP – independently of the ambient dimensions of the underlying network. We show on synthetic and real-world datasets that our method does not only achieve state-of-the-art modeling results, but also improves runtime performance by multiple orders of magnitude on sparse event sequences. In combination with easily interpretable latent variables and influence structures, this allows us to analyze diffusion processes in networks at previously unattainable scale.","PeriodicalId":106672,"journal":{"name":"Proceedings of the Web Conference 2021","volume":"101 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-04-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122891235","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}
FatemehSadat Mireshghallah, Mohammadkazem Taram, A. Jalali, Ahmed T. Elthakeb, D. Tullsen, H. Esmaeilzadeh
When receiving machine learning services from the cloud, the provider does not need to receive all features; in fact, only a subset of the features are necessary for the target prediction task. Discerning this subset is the key problem of this work. We formulate this problem as a gradient-based perturbation maximization method that discovers this subset in the input feature space with respect to the functionality of the prediction model used by the provider. After identifying the subset, our framework, Cloak, suppresses the rest of the features using utility-preserving constant values that are discovered through a separate gradient-based optimization process. We show that Cloak does not necessarily require collaboration from the service provider beyond its normal service, and can be applied in scenarios where we only have black-box access to the service provider’s model. We theoretically guarantee that Cloak’s optimizations reduce the upper bound of the Mutual Information (MI) between the data and the sifted representations that are sent out. Experimental results show that Cloak reduces the mutual information between the input and the sifted representations by 85.01% with only negligible reduction in utility (1.42%). In addition, we show that Cloak greatly diminishes adversaries’ ability to learn and infer non-conducive features.
{"title":"Not All Features Are Equal: Discovering Essential Features for Preserving Prediction Privacy","authors":"FatemehSadat Mireshghallah, Mohammadkazem Taram, A. Jalali, Ahmed T. Elthakeb, D. Tullsen, H. Esmaeilzadeh","doi":"10.1145/3442381.3449965","DOIUrl":"https://doi.org/10.1145/3442381.3449965","url":null,"abstract":"When receiving machine learning services from the cloud, the provider does not need to receive all features; in fact, only a subset of the features are necessary for the target prediction task. Discerning this subset is the key problem of this work. We formulate this problem as a gradient-based perturbation maximization method that discovers this subset in the input feature space with respect to the functionality of the prediction model used by the provider. After identifying the subset, our framework, Cloak, suppresses the rest of the features using utility-preserving constant values that are discovered through a separate gradient-based optimization process. We show that Cloak does not necessarily require collaboration from the service provider beyond its normal service, and can be applied in scenarios where we only have black-box access to the service provider’s model. We theoretically guarantee that Cloak’s optimizations reduce the upper bound of the Mutual Information (MI) between the data and the sifted representations that are sent out. Experimental results show that Cloak reduces the mutual information between the input and the sifted representations by 85.01% with only negligible reduction in utility (1.42%). In addition, we show that Cloak greatly diminishes adversaries’ ability to learn and infer non-conducive features.","PeriodicalId":106672,"journal":{"name":"Proceedings of the Web Conference 2021","volume":"22 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-04-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114280864","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}
Population health data are becoming more and more publicly available on the Internet than ever before. Such datasets offer a great potential for enabling a better understanding of the health of populations, and inform health professionals and policy makers for better resource planning, disease management and prevention across different regions. However, due to the laborious and high-cost nature of collecting such public health data, it is a common place to find many missing entries on these datasets, which challenges the utility of the data and hinders reliable analysis and understanding. To tackle this problem, this paper proposes a deep-learning-based approach, called Compressive Population Health (CPH), to infer and recover (to complete) the missing prevalence rate entries of multiple chronic diseases. The key insight of CPH relies on the combined exploitation of both intra-disease and inter-disease correlation opportunities. Specifically, we first propose a Convolutional Neural Network (CNN) based approach to extract and model both of these two types of correlations, and then adopt a Generative Adversarial Network (GAN) based prevalence inference model to jointly fuse them to facility the prevalence rates data recovery of missing entries. We extensively evaluate the inference model based on real-world public health datasets publicly available on the Web. Results show that our inference method outperforms other baseline methods in various settings and with a significantly improved accuracy (from 14.8% to 9.1%).
{"title":"Completing Missing Prevalence Rates for Multiple Chronic Diseases by Jointly Leveraging Both Intra- and Inter-Disease Population Health Data Correlations","authors":"Yujie Feng, Jiangtao Wang, Yasha Wang, A. Helal","doi":"10.1145/3442381.3449811","DOIUrl":"https://doi.org/10.1145/3442381.3449811","url":null,"abstract":"Population health data are becoming more and more publicly available on the Internet than ever before. Such datasets offer a great potential for enabling a better understanding of the health of populations, and inform health professionals and policy makers for better resource planning, disease management and prevention across different regions. However, due to the laborious and high-cost nature of collecting such public health data, it is a common place to find many missing entries on these datasets, which challenges the utility of the data and hinders reliable analysis and understanding. To tackle this problem, this paper proposes a deep-learning-based approach, called Compressive Population Health (CPH), to infer and recover (to complete) the missing prevalence rate entries of multiple chronic diseases. The key insight of CPH relies on the combined exploitation of both intra-disease and inter-disease correlation opportunities. Specifically, we first propose a Convolutional Neural Network (CNN) based approach to extract and model both of these two types of correlations, and then adopt a Generative Adversarial Network (GAN) based prevalence inference model to jointly fuse them to facility the prevalence rates data recovery of missing entries. We extensively evaluate the inference model based on real-world public health datasets publicly available on the Web. Results show that our inference method outperforms other baseline methods in various settings and with a significantly improved accuracy (from 14.8% to 9.1%).","PeriodicalId":106672,"journal":{"name":"Proceedings of the Web Conference 2021","volume":"6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-04-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129508231","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}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：