Search

With the commoditization of private data, data trading in consideration of user privacy protection has become a fascinating research topic. The trading for private web browsing histories brings huge economic value to data consumers when leveraged by targeted advertising. And the online pricing of these private data further helps achieve more realistic data trading. In this paper, we study the trading and pricing of multiple correlated queries on private web browsing history data at the same time. We propose CTRADE , which is a novel online data CommodiTization fRamework for trAding multiple correlateD queriEs over private data. CTRADE first devises a modified matrix mechanism to perturb query answers. It especially quantifies privacy loss under the relaxation of classical differential privacy and a newly devised mechanism with relaxed matrix sensitivity, and further compensates data owners for their diverse privacy losses in a satisfying manner. CTRADE then proposes an ellipsoid-based query pricing mechanism according to a given linear market value model, which exploits the features of the ellipsoid to explore and exploit the close-optimal dynamic price at each round. In particular, the proposed mechanism produces a low cumulative regret, which is quadratic in the dimension of the feature vector and logarithmic in the number of total rounds. Through real-data based experiments, our analysis and evaluation results demonstrate that CTRADE balances total error and privacy preferences well within acceptable running time, indeed produces a convergent cumulative regret with more rounds, and also achieves all desired economic properties of budget balance, individual rationality, and truthfulness.

This article argues that researchers in Translation Studies may proactively aim to understand the consequences of an envisaged merger between targeted advertising and automated translation. Functional translation software is widely available online, and several platforms now perform instant translation, sometimes without asking the user whether this is required. Indeed, the user’s main language is known to various applications, which keep track of this information along with other settings and preferences. Data tracking is commonly used to produce targeted advertising: people receive commercial information about products they are likely to be interested in. If text can instantly be altered according to a user’s linguistic preferences, it can also be altered according to aesthetic, commercial, or political preferences. The article discusses theoretical and ideological aspects of the sociotechnical evolution towards the production and consumption of personalised content, highlighting the role translation may come to play.

This work is devoted to the highly topical problem of structuring costs for contextual and targeted advertising on the Internet. The choice of the ad campaign financing structure is considered from the point of view of violating the principle of symmetry of user interest in ads. The purpose of this work is to develop a methodology for structuring advertising campaign costs based on cluster analysis, taking into account the asymmetry of user interest in advertising. The key feature of the research is the description of the possibility of using the asymmetry of user interest in application solutions, such as online advertising. The Gini coefficient is used as an indicator of the degree of imbalance in the manifestation of a feature in clustering, and the features of using the lift coefficient and the Lorentz curve to evaluate the effectiveness of contextual and targeted advertising for various groups of customers are also considered. Using the Gini index and cluster analysis, you can analyze the possibilities of increasing ad revenue and compare it with the absence of any policy for structuring advertising costs. Identifying such patterns in consumer groups allows you to identify the main directions of product development and customer interest in it. The method described here should be used to improve the effectiveness of banner advertising and clustering algorithms. This approach does not improve banner clickability, but allows you to implement an individual approach to advertising products with the current number of clicks and more effectively structure the cost of various types of advertising.

This paper addresses challenges in conceptualizing privacy posed by algorithmic systems that can infer sensitive information from seemingly innocuous data. This type of privacy is of imminent concern due to the rapid adoption of machine learning and artificial intelligence systems in virtually every industry. In this paper, we suggest informational friction, a concept from Floridi's ethics of information, as a valuable conceptual lens for studying algorithmic aspects of privacy. Informational friction describes the amount of work required for one agent to access or alter the information of another. By focusing on amount of work, rather than the type of information or manner in which it is collected, informational friction can help to explain why automated analyses should raise privacy concerns independently of, and in addition to, those associated with data collection. As a demonstration, this paper analyze law enforcement use of facial recognition, andFacebook's targeted advertising model using informational friction and demonstrate risks inherent to these systems which are not completely identified in another popular framework, Nissenbaum's Contextual Integrity.The paper concludes with a discussion of broader implications, both for privacy research and for privacy regulation.

Today, targeted advertising (TA) needs no commercial justification. Even for the linear horizontal broadcasting ecosystem, TA is a necessary evolution to be on a par with the online video distribution on all these mobile devices and personal computers (PCs) where the addressability of advertising has been a solid reality for many years. It is going to be a reality even in the “old” horizontal broadcasting ecosystems, thanks to connected TVs and recent standardization by the digital video broadcasting (DVB) Projecti and the Hybrid Broadcast Broadband TV (HbbTV) Association.ii

There have been numerous applications of artificial intelligence (AI) technologies to online advertising, especially to optimize the reach of target audiences. Previous studies show that improved computational power significantly advances granular audience targeting capabilities. This study investigates and classifies various machine learning techniques that are used to enhance targeted online advertising. Twenty-three machine learning-based online targeted advertising strategies are identified and classified largely into two categories, user-centric and content-centric approaches. The paper also identifies an underexamined area, algorithm-based detection of click frauds, to illustrate how machine learning approaches can be integrated to preserve the viability of online advertising.

Online social networks support a vibrant ecosystem of third-party apps that get access to personal information of a large number of users. Despite several recent high-profile incidents, methods to systematically detect data misuse by third-party apps on online social networks are lacking. We propose CanaryTrap to detect misuse of data shared with third-party apps. CanaryTrap associates a honeytoken to a user account and then monitors its unrecognized use via different channels after sharing it with the third-party app. We design and implement CanaryTrap to investigate misuse of data shared with third-party apps on Facebook. Specifically, we share the email address associated with a Facebook account as a honeytoken by installing a third-party app. We then monitor the received emails and use Facebook’s ad transparency tool to detect any unrecognized use of the shared honeytoken. Our deployment of CanaryTrap to monitor 1,024 Facebook apps has uncovered multiple cases of misuse of data shared with third-party apps on Facebook including ransomware, spam, and targeted advertising.

Targeted advertising is a dominant form of online advertising. It considers advertisers’ major concern of their customers, including the consumers’ certain traits, interests and individual preferences. To promote the effectiveness of advertisement delivery, advertising analysts need to understand advertiser delivery behavior and problems in targeting structure. However, statistical methods cannot meet analytical requirements completely, and analysts have to spend a lot of time reading countless data reports. Concretely, there is no efficient tool accomplishing analysis tasks such as exploring targeting usage at different levels, discovering useful or abnormal targeting combination patterns, finding competition from user behavior. In this paper, we design and implement an interactive visual analytics system named TargetingVis to visualize targeted advertising delivery data to face the challenges. After conducting a detailed requirements analysis with the domain experts from Tencent Inc., we design TargetingVis with four linked views: a novel chord diagram for cross-level exploration of targeting relations, a view for delving into the analysis of targeting combination patterns, an auxiliary view for displaying data indicators and a view to help gain insights into the behavior of advertisers. Finally, we evaluate the usability and efficiency through experiments based on real-world datasets.

Knowing accurate indoor locations of pedestrians has great social and commercial values, such as pedestrian heatmapping and targeted advertising. Location estimation with sequential inputs (e.g., geomagnetic sequences) has received much attention lately, mainly because they enhance the localization accuracy with temporal correlations. Nevertheless, it is challenging to realize accurate localization with geomagnetic sequences due to environmental factors, such as non-uniform ferromagnetic disturbances. To address this, we propose MAIL, a multi-scale attention-guided indoor localization network, which turns these challenges into favorable advantages. Our key contributions are as follows. First, instead of extracting a single holistic feature from an input sequence directly, we design a scale-based feature extraction unit that takes variational anomalies at different scales into consideration. Second, we propose an attention generation scheme that identifies attention values for different scales. Rather than setting fixed numbers, MAIL learns them adaptively with the input sequence, thus increasing its adaptability and generality. Third, guided by attention values, we fuse multi-scale features by paying more attention to prominent ones and estimate current location with the fused feature. We evaluate the performance of MAIL in three different trial sites. Evaluation results show that MAIL reduces the mean localization error by more than 36% compared with the state-of-the-art competing schemes.

The EU General Data Protection Regulation (GDPR) recognizes the data subject’s consent as one of the legal grounds for data processing. Targeted advertising, based on personal data processing, is a central source of revenue for data controllers such as Google and Facebook. At present, the implementation of consent mechanisms for such advertisements are often not well developed in practice and their compliance with the GDPR requirements can be questioned. The absence of consent may mean an unlawful data processing and a lack of control of the user (data subject) on his personal data. However, consent mechanisms that do not fully satisfy GDPR requirements can give users a false sense of control, encouraging them to allow the processing of more personal data than they would have otherwise. In this paper, we identify the features, originating from GDPR requirements, of consent mechanisms. For example, the GDPR specifies that a consent must be informed and freely given, among other requirements. We then examine the Ad Consent Mechanism of Facebook that is based on processing of user activity data off Facebook Company Products provided by third parties with respect to these features. We discuss to what extent this consent mechanism respects these features. To the best of our knowledge, our evaluation of Facebook’s Ad Consent Mechanism is the first of its kind.

With the prosperity of mobile application markets, mobile advertising is becoming an increasingly important economic force. In order to maximize revenue, ads are recommended to be delivered to potentially interested users, which requires user targeting, i.e., analyzing users’ profiles and exploring users’ interests. However, collecting user personal information for targeted mobile advertising services raises critical privacy concerns. Although some solutions like anonymization and obfuscation have been proposed for privacy-aware targeted advertising, they undesirably face the issues of security, efficiency, and/or ad relevance. In this paper, we propose a practical system enabling secure and efficient targeted mobile advertising services. It allows the ad network to perform accurate user targeting, while ensuring strong privacy protection for mobile users. Specifically, we show how to properly leverage a cryptographic primitive called private stream searching to support secure, accurate, and practical targeted mobile ad delivery. Moreover, we propose secure billing schemes to enable the ad network to charge advertisers in a privacy-preserving manner. The security strength of our system is thoroughly analyzed. Through extensive experiments, we show that our system achieves practical efficiency on mobile devices.

We consider an online advertisement system and focus on the impact of user interaction and response to targeted advertising campaigns. We analytically model the system dynamics accounting for the user behavior and devise strategies to maximize a relevant metric called click-through-intensity (CTI), defined as the number of clicks per time unit. With respect to the traditional click-through-rate (CTR) metric, CTI better captures the success of advertisements for services that the users may access several times, making multiple purchases or subscriptions. Examples include advertising of on-line games or airplane tickets. The model we develop is validated through traces of real advertising systems and allows us to optimize CTI under different scenarios depending on the nature of ad delivery and of the information available at the system. Experimental results show that our approach can increase the revenue of an ad campaign, even when user’s behavior can only be estimated.

Over the past decade, keystroke-based pattern recognition techniques, as a forensic tool for behavioral biometrics, have gained increasing attention. Although a num- ber of machine learning-based approaches have been proposed, they are limited in terms of their capability to recognize and profile a set of an individual’s characteristics. In addition, up to today, their focus was primarily gender and age, which seem to be more appropriate for commercial applications (such as developing commercial software), leaving out from research other characteristics, such as the educational level. Educational level is an acquired user characteristic, which can improve targeted advertising, as well as provide valuable information in a digital forensic investigation, when it is known. In this context, this paper proposes a novel machine learning model, the randomized radial basis function network, which recognizes and profiles the edu- cational level of an individual who stands behind the keyboard. The performance of the proposed model is evaluated by using the empirical data obtained by recording volunteers’ keystrokes during their daily usage of a computer. Its performance is also compared with other well-referenced machine learning models using our keystroke dynamic datasets. Although the proposed model achieves high accuracy in educational level prediction of an unknown user, it suffers from high computational cost. For this reason, we examine ways to reduce the time that is needed to build our model, including the use of a novel data condensation method, and discuss the tradeoff between an accurate and a fast prediction. To the best of our knowledge, this is the first model in the literature that predicts the educational level of an individual based on the keystroke dynamics information only.

Recommendation systems form the centerpiece of a rapidly growing trillion dollar online advertisement industry. Curating and storing profile information of users on web portals can seriously breach...

Digital signage is an important outdoor advertising medium in cities. However, advertising on digital signage often lacks pertinence. Thus, it is important to introduce an accurate digital signage audience classification method to facilitate targeted advertising. In this study, a multi-label classification model based on a backpropagation (BP) neural network and the Huff model, referred to as the Huff-BP model, is proposed to investigate digital signage audience classification. A case study is performed on outdoor digital signage within the 6th Ring Road in Beijing, China, and economic census, population census, average housing price, social media check-in and the centrality of traffic networks as research data. The data are divided into 100 × 100-1,000 × 1,000 m normal grids. Multi-label classification modelling factors for various grid scales are constructed. The BP neural network classification algorithm is improved to solve the multilabel classification problem. In addition, an improved Huff model is used to calculate the digital signage influence values between each grid cell and integrated into the improved BP neural network to classify modelling factors at various scales. Finally, four metrics are used to examine the effectiveness of the proposed model. The results show that the Huff-BP-based multi-label classification model achieves relatively good classification results, and the digital signage audiences are mainly concentrated within the 4th Ring Road and near the 5th Ring Road.

The era of E-commerce and availability of data in every field of operations in an enormous volume that implies to Big Data is one of the biggest sources of competitive advantage for the organizations in this digital world. It provides useful information to grow businesses by posting the advertisement and help consumers find the relevant product according to their preferences. The focus of this research is on the advertisement strategies analysis under which a Business employs several online advertisement strategies in order to appeal to the consumer. This research work will present a detailed analysis in user behavioral to use for business or Online Behavioral advertising and provide the framework of how Enterprise Resource Planning systems track the targeted audience and show their content. The paper's prime objective is to classify and effectively run targeted advertising using the data that shows user's retail behavior. This is where an Enterprise Resource Planning driven data will give rise to behavioral analytics. In addition to this, various data streaming technologies are also emphasized that will help to create a pipeline for the huge amount of data in Enterprise Resource Planning's database.

The Analytics companies enable successful targeted advertising via user profiles, derived from the mobile apps installed by specific users, and hence have become an integral part of the mobile advertising industry. This threatens the users' privacy, when profiling is based on apps representing sensitive information, e.g., gambling problems indicated by a game app. In this work, we propose an app-based profile obfuscation mechanism, ProfileGuard, with the objective of eliminating the dominance of private interest categories (i.e. the prevailing private interest categories present in a user profile). We demonstrate, based on wide-range experimental evaluation of Android apps in a nine month test campaign, that the proposed obfuscation mechanism based on similarity with user's existing apps (ensuring that selected obfuscating apps belong to non-private categories) can achieve a good trade-off between efforts required by the obfuscating system and the resulting privacy protection. We also show how the bespoke (customised to profile obfuscation) and bespoke++ (resource-aware) strategies can deliver significant improvements in the level of obfuscation and (particularly bespoke++) in the use of mobile resources, making the latter a good candidate strategy in resource-constrained scenarios e.g., for fixed data use mobile plans. We also implement a POC ProfileGuard app to demonstrate the feasibility of an automated obfuscation mechanism. Furthermore, we provide insights to Google AdMob profiling rules, such as showing how individual apps map to user's interests within their profile in a deterministic way and that AdMob requires a certain level of activity to build a stable user profile.

Mobile digital billboards are an effective way to augment brand-awareness. Among various such mobile billboards, taxicab rooftop devices are emerging in the market as a brand new media. Motov is a leading company in South Korea in the taxicab rooftop advertising market. In this work, we present a lightweight yet accurate deep learning-based method to predict taxicabs' next locations to better prepare for targeted advertising based on demographic information of locations. Considering the fact that next POI recommendation datasets are frequently sparse, we design our presented model based on neural ordinary differential equations (NODEs), which are known to be robust to sparse/incorrect input, with several enhancements. Our model, which we call LightMove, has a larger prediction accuracy, a smaller number of parameters, and/or a smaller training/inference time, when evaluating with various datasets, in comparison with state-of-the-art models.

Targeted or “personalized” marketing is an everyday part of most web users’ experience. But how do companies “personalize” commercial web content in the context of mass data aggregation? What does it really mean to use data to target web users by their “personal interests” and individual identities? What kinds of ethical implications arise from such practices? This case study explores commercial algorithmic profiling, targeting, and advertising systems, considering the extent to which such systems can be ethical. To do so the case study first maps a brief history of the commercially targeted user, then explores how web users themselves perceive targeted advertising in relation to data knowledge, cookie consent, and “algorithmic disillusionment.” It goes on to analyze current regulatory landscapes and consider how developers who target audiences might avoid placing burdens of impossible data choice on web users themselves. Finally, it offers a series of reflections on best practice in terms of how (not) to profile and target web users. To illuminate the ethical considerations connected to commercial targeted advertising systems, this case study presents some study tasks (see Exercises 1 and 2) that can be used as discussion points for those interested in exploring the nuances of targeting in specific contexts.

Automated influence, delivered by digital targeting technologies such as targeted advertising, digital nudges, and recommender systems, has attracted significant interest from both empirical researchers, on one hand, and critical scholars and policymakers on the other. In this paper, we argue for closer integration of these efforts. Critical scholars and policymakers, who focus primarily on the social, ethical, and political effects of these technologies, need empirical evidence to substantiate and motivate their concerns. However, existing empirical research investigating the effectiveness of these technologies (or lack thereof), neglects other morally relevant effects-which can be felt regardless of whether or not the technologies "work" in the sense of fulfilling the promises of their designers. Drawing from the ethics and policy literature, we enumerate a range of questions begging for empirical analysis-the outline of a research agenda bridging these fields---and issue a call to action for more empirical research that takes these urgent ethics and policy questions as their starting point.

With the explosion of multimedia data, machine learning applications are proliferating. Privacy concerns have recently resurfaced, due to a few prominent incidents of user data leakage, which divulged sensitive information to other people. Multimedia data platform providers––social networks, their affiliates or governments with access to users’ content––use algorithms to profile users by extracting or inferring demographic information, personality traits, relationships, opinions, and beliefs. These results, in turn, feed algorithms for targeted advertising but also for personalized content recommendation to maximize user engagement. While this is ostensibly for users’ benefit, they have serious privacy implications. We highlight this new problem of privacy protection against machines in contrast to the traditional problem of privacy protection against humans. We briefly touch upon our initial solution, a human-sensitivity-aware image perturbation model, which is able to modify the computational classification results of sensitive attributes while preserving the remaining attributes. We then point to many exciting open problems in this new area.

Microblogs data is the microlength user-generated data that is posted on the web, e.g., tweets, online reviews, comments on news and social media. It has gained considerable attention in recent years due to its widespread popularity, rich content, and value in several societal applications. Nowadays, microblogs applications span a wide spectrum of interests including targeted advertising, market reports, news delivery, political campaigns, rescue services, and public health. Consequently, major research efforts have been spent to manage, analyze, and visualize microblogs to support different applications. This paper gives a comprehensive review of major research and system work in microblogs data management. The paper reviews core components that enable large-scale querying and indexing for microblogs data. A dedicated part gives particular focus for discussing system-level issues and on-going effort on supporting microblogs through the rising wave of big data systems. In addition, we review the major research topics that exploit these core data management components to provide innovative and effective analysis and visualization for microblogs, such as event detection, recommendations, automatic geotagging, and user queries. Throughout the different parts, we highlight the challenges, innovations, and future opportunities in microblogs data research.

Due to the rapid proliferation in the algorithm-driven online tracking and profiling infrastructure, and the increasing business potential of online behavioral advertising, researchers from...

International audience; The paper focuses on the analysis of multi-agent systems interacting over directed and time-varying networks in presence of parametric uncertainty on the interaction weights. We assume that agents reach a consensus and the main goal of this work is to characterize the contribution that each agent has to the consensus value. This information is important for network intervention applications such as targeted advertising over social networks. Indeed, for an advertising campaign to be efficient, it has to take into account the influence power of each agent in the graph (i.e., the contribution of each agent to the final consensus value). In our first results we analytically describe the trajectory of the overall network and we provide lower and upper bounds on the corresponding consensus value. We show that under appropriate assumptions, the contribution of each agent to the consensus value is smooth both in time and in the variation of the uncertainty parameter. This allows approximating the contribution of each agent when small perturbations affect the influence of each agent on its neighbors. Finally, we provide a numerical example to illustrate how our theoretical results apply in the context of network intervention.

In this paper, we present a method to identify forum posts expressing user intentions in online discussion forums. The results of this task, for example buying intentions, can be exploited for targeted advertising or other marketing tasks. Our method utilizes labeled data from other domains to help the learning task in the target domain by using a Naive Bayes (NB) framework to combine the data statistics . Because the distributions of data vary from domain to domain, it is important to adjust the contributions of different data sources when constructing the learning model, to achieve accurate results. Here, we propose to adjust the parameters of the NB classifier by optimizing an objective, which is equivalent to maximizing the between-class separation, using stochastic gradient descent. Experimental results show that our method outperforms several competitive baselines on a benchmark dataset consisting of forum posts from four domains: Cellphone, Electronics, Camera, and TV. In addition, we explore the possibility of combining NB posteriors computed during the optimization process with another classifier, namely Support Vector Machines. Experimental results show the usefulness of optimized NB class posteriors when using as features for SVMs in the cross-domain settings.

Driven by the advance of positioning technology and the tremendous popularity of location-based services, location-record data have become unprecedentedly available. Publishing such data is of vital importance to the advancement of a wide spectrum of applications, such as marketing analysis, targeted advertising, and urban planning. However, the data collection may pose considerable threats to the individuals privacy. Local differential privacy (LDP) has recently emerged as a strong privacy standard for collecting sensitive information from users. Due to the inherent high dimensionality, it is particularly challenging to publish the location-record data under LDP. In this paper, we propose LDPart , a probabilistic top-down partitioning algorithm to effectively generate a sanitized location-record data. Our approach employs a carefully designed partition tree model to extract essential information in terms of location records. Furthermore, it also makes use of a novel adaptive user allocation scheme and a series of optimization techniques to improve the accuracy of the released data. The extensive experiments conducted on real-world datasets demonstrate that the proposed approach maintains high utility while providing privacy guarantees.

We study identifying user clusters in contextual multi-armed bandits (MAB). Contextual MAB is an effective tool for many real applications, such as content recommendation and online advertisement. In practice, user dependency plays an essential role in the user's actions, and thus the rewards. Clustering similar users can improve the quality of reward estimation, which in turn leads to more effective content recommendation and targeted advertising. Different from traditional clustering settings, we cluster users based on the unknown bandit parameters, which will be estimated incrementally. In particular, we define the problem of cluster detection in contextual MAB, and propose a bandit algorithm, LOCB, embedded with local clustering procedure. And, we provide theoretical analysis about LOCB in terms of the correctness and efficiency of clustering and its regret bound. Finally, we evaluate the proposed algorithm from various aspects, which outperforms state-of-the-art baselines., 13 pages

Ad platforms such as Facebook, Google and LinkedIn promise value for advertisers through their targeted advertising. However, multiple studies have shown that ad delivery on such platforms can be skewed by gender or race due to hidden algorithmic optimization by the platforms, even when not requested by the advertisers. Building on prior work measuring skew in ad delivery, we develop a new methodology for black-box auditing of algorithms for discrimination in the delivery of job advertisements. Our first contribution is to identify the distinction between skew in ad delivery due to protected categories such as gender or race, from skew due to differences in qualification among people in the targeted audience. This distinction is important in U.S. law, where ads may be targeted based on qualifications, but not on protected categories. Second, we develop an auditing methodology that distinguishes between skew explainable by differences in qualifications from other factors, such as the ad platform's optimization for engagement or training its algorithms on biased data. Our method controls for job qualification by comparing ad delivery of two concurrent ads for similar jobs, but for a pair of companies with different de facto gender distributions of employees. We describe the careful statistical tests that establish evidence of non-qualification skew in the results. Third, we apply our proposed methodology to two prominent targeted advertising platforms for job ads: Facebook and LinkedIn. We confirm skew by gender in ad delivery on Facebook, and show that it cannot be justified by differences in qualifications. We fail to find skew in ad delivery on LinkedIn. Finally, we suggest improvements to ad platform practices that could make external auditing of their algorithms in the public interest more feasible and accurate., To appear in The Web Conference (WWW 2021)

Part 6: Potpourri II; International audience; We present a Divide-and-Learn machine learning methodology to investigate a new class of attribute inference attacks against Online Social Networks (OSN) users. Our methodology analyzes commenters' preferences related to some user publications (e.g., posts or pictures) to infer sensitive attributes of that user. For classification performance, we tune Random Indexing (RI) to compute several embeddings for textual units (e.g., word, emoji), each one depending on a specific attribute value. RI guarantees the comparability of the generated vectors for the different values. To validate the approach, we consider three Facebook attributes: gender, age category and relationship status, which are highly relevant for targeted advertising or privacy threatening applications. By using an XGBoost classifier, we show that we can infer Facebook users' attributes from commenters' reactions to their publications with AUC from 94% to 98%, depending on the traits.

Machine translation of user-generated code-mixed inputs to English is of crucial importance in applications like web search and targeted advertising. We address the scarcity of parallel training data for training such models by designing a strategy of converting existing non-code-mixed parallel data sources to code-mixed parallel data. We present an m-BERT based procedure whose core learnable component is a ternary sequence labeling model, that can be trained with a limited code-mixed corpus alone. We show a 5.8 point increase in BLEU on heavily code-mixed sentences by training a translation model using our data augmentation strategy on an Hindi-English code-mixed translation task.

Despite its core role in the EU system of data protection, the notion of identification has been a neglected subject in the data protection law and scholarship. With a spotlight focused on the meaning of identifiability as a legally relevant possibility of identification, it remained unclear the possibility of what exactly is at issue. While Article 29 Working Party interpreted identification broadly, as distinguishing one in a group, this interpretation has been questioned in light of the EUCJ decision in Breyer, and the uncertainty remained. The issue tackled in this paper is the meaning of identification under the GDPR. The contribution of the paper is three-fold. First, it offers an integrated typology of identification outside of the legal context. The typology builds on three socio-technical accounts of identification: four identifiability types by Leenes, seven types of identity knowledge by Marx, and anonymity as unreachability by Nissenbaum. Second, it identifies personalisation as a new identification type, i.e. a relatively unique characterization, where one is individualized by being mapped in relation to multiple dimensions within a multidimensional space. The argument builds on the literatures on calculated publics, profiling in recommender systems, agile methods of software development, price and content personalization. Third, it clarifies the meaning of identification under the GDPR. I propose a contextual interpretation of Breyer, which negates Breyer’s restrictive potential and brings all identification types within the GDPR. The paper discusses implications of this reading of identification for data protection law and research.

As an emerging social dynamic system, geo-social network can be used to facilitate viral marketing through the wide spread of targeted advertising. However, unlike traditional influence spread problem, the heterogeneous spatial distribution has to incorporated into geo-social network environment. Moreover, from the perspective of business managers, it is indispensable to balance the tradeoff between the objective of influence spread maximization and objective of promotion cost minimization. Therefore, these two goals need to be seamlessly combined and optimized jointly. In this paper, considering the requirements of real-world applications, we develop a multiobjective optimization-based influence spread framework for geo-social networks, revealing the full view of Pareto-optimal solutions for decision makers. Based on the reverse influence sampling (RIS) model, we propose a similarity matching-based RIS sampling method to accommodate diverse users, and then transform our original problem into a weighted coverage problem. Subsequently, to solve this problem, we propose a greedy-based incrementally approximation approach and heuristic-based particle swarm optimization approach. Extensive experiments on two real-world geo-social networks clearly validate the effectiveness and efficiency of our proposed approaches.

Consider a publisher of online display advertising that sells its ad resources in both an upfront market and a spot market. When planning its ad delivery, the publisher needs to make a trade-off between earning a greater short-term profit from the spot market and improving advertising effectiveness in the upfront market. To address this challenge, we propose an integrated planning model that is robust to the uncertainties associated with the supply of advertising resources. Specifically, we model the problem as a distributionally robust chance-constrained program. We first approximate the program by using a robust optimization model, which is then transformed into a linear program. We provide a theoretical bound on the performance loss due to this transformation. A clustering algorithm is proposed to solve large-scale cases in practice. We implement ad serving of our planning model on two real data sets, and we demonstrate how to incorporate realistic constraints such as exclusivity and frequency caps. Our numerical experiments demonstrate that our approach is very effective: it generates more revenue while fulfilling the guaranteed contracts and ensuring advertising effectiveness.

Today deep learning has provided us with endless possibilities for solving problems in many domains. Diagnosing diseases, speech recognition, image classification, and targeted advertising are a few of its applications. Starting this process from scratch requires using large amounts of labeled data and significant cloud processing usage. Transfer learning is a deep learning technique that solves this problem by making use of a model that is pre-trained for a certain task and using it on a different task of a related problem. Therefore, the goal of the project is to utilize transfer learning and achieve near-perfect results using a limited amount of data and computation power. To demonstrate, an image classifier using FastAI that detects three types of birds with up to 94% accuracy is implemented. This approach can be applied to solve tasks that are limited by labeled data and would gain by knowledge learned from a related task.

Social services used for information storage, communication, marketing, and news such as Facebook, Instagram, and Twitter utilize big data. This data is often used by researchers and marketing companies in systems such as targeted advertising and product recommendations. In this study, we analyze the text information posted on Twitter to obtain new marketing data and compare algorithms for estimating the user's personal interests. Specifically, we used the Twitter API to retrieve text information and pre-process that text to remove noise. Then we estimate hobby classes comparing three algorithm methods: naive Bayesian (NB), support vector machine (SVM), and multilayer perceptions (MLP).As a result, the average accuracies for each algorithm were 0.81 for the naive Bayesian method and 0.84 for both SVM and MLP. There was no difference between SVM and MLP, although SVM and MLP had a higher score of 0.84.

Domain classification services have applications in multiple areas, including cybersecurity, content blocking, and targeted advertising. Yet, these services are often a black box in terms of their methodology to classifying domains, which makes it difficult to assess their strengths, aptness for specific applications, and limitations. In this work, we perform a large-scale analysis of 13 popular domain classification services on more than 4.4M hostnames. Our study empirically explores their methodologies, scalability limitations, label constellations, and their suitability to academic research as well as other practical applications such as content filtering. We find that the coverage varies enormously across providers, ranging from over 90% to below 1%. All services deviate from their documented taxonomy, hampering sound usage for research. Further, labels are highly inconsistent across providers, who show little agreement over domains, making it difficult to compare or combine these services. We also show how the dynamics of crowd-sourced efforts may be obstructed by scalability and coverage aspects as well as subjective disagreements among human labelers. Finally, through case studies, we showcase that most services are not fit for detecting specialized content for research or content-blocking purposes. We conclude with actionable recommendations on their usage based on our empirical insights and experience. Particularly, we focus on how users should handle the significant disparities observed across services both in technical solutions and in research. Tim Burke (author no. 6) is NOT u0109837, but rather r0304836 (a student) ispartof: pages:598-618 ispartof: Proceedings of the 2020 ACM Internet Measurement Conference pages:598-618 ispartof: 2020 ACM Internet Measurement Conference location:Virtual Event, USA date:27 Oct - 29 Oct 2020 status: Published online

The success of platforms such as Facebook and Google has been due in no small part to features that allow advertisers to target ads in a fine-grained manner. However, these features open up the potential for discriminatory advertising when advertisers include or exclude users of protected classes---either directly or indirectly---in a discriminatory fashion. Despite the fact that advertisers are able to compose various targeting features together, the existing mitigations to discriminatory targeting have focused only on individual features; there are concerns that such composition could result in targeting that is more discriminatory than the features individually.In this paper, we first demonstrate how compositions of individual targeting features can yield discriminatory ad targeting even for Facebook's restricted targeting features for ads in special categories (meant to protect against discriminatory advertising). We then conduct the first study of the potential for discrimination that spans across three major advertising platforms (Facebook, Google, and LinkedIn), showing how the potential for discriminatory advertising is pervasive across these platforms. Our work further points to the need for more careful mitigations to address the issue of discriminatory ad targeting.

We consider the task of selection personalized advertisement for Internet users in the targeted advertising system. In fact, this leads to the regression problem, when for an arbitrary user U, it is necessary to predict the click probability on a set of banners B1…Bn in order to select the most suitable banners. The real values of the predicted probabilities are also important because they may be used in an auction between different advertising systems on many sites. Since the users’ interests and the set of banners are often changed, it is necessary to train the model in online mode. In addition, large advertising systems have to deal with a large amount of data that needs to be processed in real-time. This limits the complexity of the applicable models. Therefore, linear models that are well suited for dynamic learning remain popular for this task. However, data are rarely linearly separable, and therefore, when using such models, it is required to construct derivative features, for example, by hashing combinations of the original features. A serious drawback is that these combinations are needed to select manually. In this paper, it is proposed to use a neural network with specialized architecture to avoid this problem. Special attention is paid to the analysis of the results on the test set, for which a specialized statistical testing technique is used. The results of testing showed that a neural network model with automatically constructed features works equally with logistic regression with manually selected combinations for hashing.

Targeting system is an essential part of computational advertising. It allows advertisers to select and reach their targeted users. Due to various advertising goals and the demand for making budget plans, advertisers have a strong will to configure the final targeting results, or they can become very cautious in spending money on advertising campaigns. Meanwhile, to guarantee the advertising performance, the targeted users should also be relevant to the ads of the advertisers. Recent targeting methods are mainly based on tags produced by the Data Management Platform (DMP) which is easy for the advertisers to configure the targeting results. However, in such methods, the relevance between the targeted users and ads is not technically evaluated and cannot be guaranteed. The biggest challenge is that it is hard for a machine learning model to both model the relevance and take account of the advertiser’s configuration demands. In this paper, we propose a novel relevance-driven and configurable targeting framework called Smart Targeting to solve the problem. Specifically, different from Tag-wise Targeting, we first use a relevance model to retrieve the most relevant users for the ads. To further enable the advertisers to configure the final results, we develop a Delay Intervention Mechanism to leverage the power of DMP. As far as we know, this is the first attempt of combining relevance modeling and advertiser intervention into a unified targeting system. We implement and evaluate our framework on JD.com platform with over 300 million users and the results show that it can bring significant improvements to the core indicators such as CTR and eCPM. The long term monitoring also demonstrates that Smart Targeting gradually becomes the most popular targeting tool after its release.

International audience; We present an online attribute inference attack by leverag-ing Facebook picture metadata (i) alt-text generated by Facebook to describe picture contents, and (ii) comments containing words and emo-jis posted by other Facebook users. Specifically, we study the correlation of the picture's owner with Facebook generated alt-text and comments used by commenters when reacting to the image. We concentrate on gender attribute that is highly relevant for targeted advertising or privacy breaking. We explore how to launch an online gender inference attack on any Facebook user by handling online newly discovered vocabulary using the retrofitting process to enrich a core vocabulary built during offline training. Our experiments show that even when the user hides most public data (e.g., friend list, attribute, page, group), an attacker can detect user gender with AUC (area under the ROC curve) from 87% to 92%, depending on the picture metadata availability. Moreover, we can detect with high accuracy sequences of words leading to gender disclosure, and accordingly, enable users to derive countermeasures and configure their privacy settings safely.

The targeted advertising is based on preference profiles inferred via relationships among individuals, their monitored responses to previous advertising and temporal activity over the Internet, which has raised critical privacy concerns. In this paper, we present a novel proposal for a Blockchain-based advertising platform that provides: a system for privacy preserving user profiling, privately requesting ads from the advertising system, the billing mechanisms for presented and clicked ads, the advertising system that uploads ads to the cloud according to profiling interests, various types of transactions to enable advertising operations in Blockchain-based network, and the method that allows a cloud system to privately compute the access policies for various resources (such as ads, mobile user profiles). Our main goal is to design a decentralized framework for targeted ads, which enables private delivery of ads to users whose behavioral profiles accurately match the presented ads, defined by the ad system. We implement a POC of our proposed framework i.e. a Bespoke Miner and experimentally evaluate various components of Blockchain-based in-app advertising system, implementing various critical components; such as, evaluating user profiles, implementing access policies, encryption and decryption of users' profiles. We observe that the processing delay for traversing policies of various tree sizes, the encryption/decryption time of user profiling with various key-sizes and user profiles of various interests evaluates to an acceptable amount of processing time as that of the currently implemented ad systems.

With the proliferation of electronic commerce (e-commerce), the data generated by both customers and service providers can accumulate at a fast rate. As such, analyzing the rich but subtle patterns within the e-commerce data offers a prominent opportunity of refining user experience and increasing business revenue. Due to the high velocity of e-commerce data, sequence modelling plays a pivotal role in delivering timely predictive analytics and recommendations. Based on the granularity of data, sequence modelling for e-commerce is mainly conducted at two levels, namely macro-level modelling and micro-level modelling. When researching on e-commerce data, macro-level sequence modelling aims to understand the evolution of high-level business trends in order to set the foundation for enterprise strategic planning, e.g., sales prediction for inventory management. Meanwhile, micro-level sequence modelling focuses on learning fine-grained and dynamic user preferences from behavioral data to deliver personalized user experience, e.g., recommendation systems deployed by all major e-commerce platforms. In our research, we aim to effectively tackle sequence modelling in e-commerce scenarios at different levels, and then propose a unified model that allows for both macro- and micro-level sequence modelling, thus supporting a wide range of e-commerce applications. In summary, our research consists of the following three parts.Firstly, for macro-level sequence modelling, we solve the problem of sales prediction, which is a critical means to achieve a healthy balance between supply and demand in e-commerce. The sales prediction task is formulated as a time series prediction problem which aims to predict the future sales volume for different products with observed influential factors (e.g., brand, season, discount, etc.) and corresponding historical sales records. However, with the development of contemporary commercial markets, the dynamic interactions between influential factors with different semantic meanings become more subtle, causing challenges in fully capturing dependencies among these variables. Besides, though seeking similar trends from the history benefits the accuracy for the prediction, existing methods hardly suit sales prediction tasks because the trends in sales data are more irregular and complex. Hence, we gain insights from the encoder-decoder recurrent neural network (RNN) structure, and propose a novel framework named TADA to carry out trend alignment with dual-attention, multi-task RNNs for sales prediction. In TADA, we innovatively divide the influential factors into internal feature and external feature, which are jointly modelled by a multi-task RNN encoder. In the decoding stage, TADA utilizes two attention mechanisms to compensate for the unknown states of influential factors in the future and adaptively align the upcoming trend with relevant historical trends to ensure precise sales prediction.nSecondly, for micro-level sequence modelling, we investigate sequential top-k recommendation, which infers users' preferences from their sequential behaviors and predicts their next interested items. Though it is important to capture the sequential patterns from the user-item interaction data, existing methods only focus on modelling the sparse item-wise sequential effect in user preference and only consider the homogeneous user interaction behaviors (i.e., a single type of user behavior). As a result, the data sparsity issue inevitably arises and makes the learned sequential patterns fragile and unreliable, impeding the sequential recommendation performance of existing methods. Hence, in this task, we propose AIR, namely attentional intention-aware recommender systems to predict category-wise future user intention and collectively exploit the rich heterogeneous user interaction behaviors (i.e., multiple types of user behaviors). In AIR, we propose to represent user intention as an action-category tuple to discover category-wise sequential patterns and to capture varied effect of different types of actions for recommendation. A novel attentional recurrent neural network (ARNN) is proposed to model the intention migration effect and infer users' future intention. Besides, an intention-aware factorization machine (ITFM) is developed to perform intention-aware sequential recommendation.Lastly, we develop a machine learning model that is generalizable to both macro- and micro-level sequence modelling tasks in e-commerce. Specifically, we extend a versatile predictive model, namely factorization machines (FMs) to the sequential setting. In e-commerce, models based on FMs are capable of modelling high-order interactions among features for effective predictive analytics, e.g., targeted advertising and recommendation. However, existing FM-based models assume no temporal orders in the data, and are unable to capture the sequential dependencies or patterns within the dynamic features, impeding the performance and adaptivity of these methods. Hence, we propose a novel sequence-aware factorization machine (SeqFM) for sequential predictive analytics, which models feature interactions by fully investigating the effect of sequential dependencies. As static features (e.g., user gender) and dynamic features (e.g., users' interacted items) express different semantics, we innovatively devise a multi-view self-attention scheme that separately models the effect of static features, dynamic features and the mutual interactions between static and dynamic features in three different views. In SeqFM, we further map the learned representations of feature interactions to the desired output with a shared residual network.n

With the commoditization of private data, data trading in consideration of user privacy protection has become a fascinating research topic. The trading for private web browsing histories brings huge economic value to data consumers when leveraged by targeted advertising. In this paper, we study the trading of multiple correlated queries on private web browsing history data. We propose TERBE, which is a novel trading framework for correlaTed quEries based on pRivate web Browsing historiEs. TERBE first devises a modified matrix mechanism to perturb query answers. It then quantifies privacy loss under the relaxation of classical differential privacy and a newly devised mechanism with relaxed matrix sensitivity, and further compensates data owners for their diverse privacy losses in a satisfying manner. Through real-data based experiments, our analysis and evaluation results demonstrate that TERBE balances total error and privacy preferences well within acceptable running time, and also achieves all desired economic properties of budget balance, individual rationality, and truthfulness.

In an online social network, users exhibit personal information to enjoy social interaction. The social network provider (SNP) exploits users’ information for revenue generation through targeted advertising. The SNP can present ads to proper users efficiently. Therefore, an advertiser is more willing to pay for targeted advertising. However, the over-exploitation of users’ information would invade users’ privacy, which would negatively impact users’ social activeness. Motivated by this, we study the optimal privacy policy of the SNP with targeted advertising business. We characterize the privacy policy in terms of the fraction of users’ information that the provider should exploit, and formulate the interactions among users, advertiser, and SNP as a three-stage Stackelberg game. By carefully leveraging supermodularity property, we reveal from the equilibrium analysis that higher information exploitation will discourage users from exhibiting information, lowering the overall amount of exploited information and harming advertising revenue. We further characterize the optimal privacy policy based on the connection between users’ information levels and privacy policy. Numerical results reveal some useful insights that the optimal policy can well balance the users’ trade-off between social benefit and privacy loss.

Smartphones carry a plethora of sensitive and personally identifiable information (PII) such as email addresses, GPS coordinates, names, and phone numbers. A common occurrence in the design of many popular smartphone applications is to harvest this user data for consumer market analysis and targeted advertising. Transmitting sensitive PII data without the user’s explicit knowledge has been given the name “information leakage Unfortunately, the permission systems employed by modern smartphone OSes are too coarse grained, presenting an “all or nothing” choice to users making it largely insufficient to defend against information leakage attacks. In this paper we propose a network-filtering based solution, which uses an entirely on-device VPN to capture and scan network packets for PII data. Our novelty is a specially designed string searching algorithm used to scan network packets, and a Naive Bayes classifier to learn and predict the user’s desired action when information leakage occurs. We evaluate and compare our work to other recent literature. We achieve 2MB/s throughput with our string searching algorithm and ~66% accuracy with our Naive Bayes classifier after building a training set of only 50 observations.

With the development of online advertising technology, the accurate targeted advertising based on user preferences is obviously more suitable both for the market and users. The amount of conversion can be properly increased by predicting the user&rsquo, s purchasing intention based on the advertising Conversion Rate (CVR). According to the high-dimensional and sparse characteristics of the historical behavior sequences, this paper proposes a LSLM_LSTM model, which is for the advertising CVR prediction based on large-scale sparse data. This model aims at minimizing the loss, utilizing the Adaptive Moment Estimation (Adam) optimization algorithm to mine the nonlinear patterns hidden in the data automatically. Through the experimental comparison with a variety of typical CVR prediction models, it is found that the proposed LSLM_LSTM model can utilize the time series characteristics of user behavior sequences more effectively, as well as mine the potential relationship hidden in the features, which brings higher accuracy and trains faster compared to those with consideration of only low or high order features.