WWW 2025 论文列表

While advertising has become commonplace in today's online interactions, there is a notable dearth of research investigating the extent to which browser fingerprinting is harnessed for user tracking and targeted advertising. Prior studies only measured whether fingerprinting-related scripts are being run on the websites but that in itself \textit{does not} necessarily mean that fingerprinting is being used for the privacy-invasive purpose of online tracking because fingerprinting might be deployed for the defensive purposes of bot/fraud detection and user authentication. It is imperative to address the mounting concerns regarding the utilization of browser fingerprinting in the realm of online advertising. To understand the privacy-invasive use of fingerprinting for user tracking, this paper introduces a new framework ``FPTrace'' (fingerprinting-based tracking assessment and comprehensive evaluation framework) designed to identify alterations in advertisements resulting from adjustments in browser fingerprinting settings. Our approach involves emulating genuine user interactions, capturing advertiser bid data, and closely monitoring HTTP information. Using FPTrace, we conduct a large scale measurement study to identify whether browser fingerprinting is being used for the purpose of user tracking and ad targeting. The results we have obtained provide robust evidence supporting the utilization of browser fingerprinting for the purposes of advertisement tracking and targeting. This is substantiated by significant disparities in bid values and a reduction in HTTP records subsequent to changes in fingerprinting. %Additionally, our study delved into the impact of browser fingerprinting on the restoration of cookies, and no conclusive evidence to support browser fingerprinting's direct involvement in cookie restoration. We additionally demonstrate the potential use of fingerprinting for privacy-evading online tracking purposes even when users opt out of tracking under GDPR/CCPA regulations. In conclusion, our research unveils the widespread employment of browser fingerprinting in online advertising, prompting critical considerations regarding user privacy and data security within the digital advertising landscape.

Multi-criteria (MC) recommender systems, which utilize MC rating information for recommendation, are increasingly widespread in various e-commerce domains. However, the MC recommendation using training-based collaborative filtering, requiring consideration of multiple ratings compared to single-criterion counterparts, often poses practical challenges in achieving state-of-the-art performance along with scalable model training. To solve this problem, we propose CA-GF, a training-free MC recommendation method, which is built upon criteria-aware graph filtering for efficient yet accurate MC recommendations. Specifically, first, we construct an item--item similarity graph using an MC user-expansion graph. Next, we design CA-GF composed of the following key components, including 1) criterion-specific graph filtering where the optimal filter for each criterion is found using various types of polynomial low-pass filters and 2) criteria preference-infused aggregation where the smoothed signals from each criterion are aggregated. We demonstrate that CA-GF is (a) efficient: providing the computational efficiency, offering the extremely fast runtime of less than 0.2 seconds even on the largest benchmark dataset, (b) accurate: outperforming benchmark MC recommendation methods, achieving substantial accuracy gains up to 24% compared to the best competitor, and (c) interpretable: providing interpretations for the contribution of each criterion to the model prediction based on visualizations.

Social media platforms employ various content moderation techniques to remove harmful, offensive, and hate speech content. The moderation level varies across platforms; even over time, it can evolve in a platform. For example, Parler, a fringe social media platform popular among conservative users, was known to have the least restrictive moderation policies, claiming to have open discussion spaces for their users. However, after linking the 2021 US Capitol Riots and the activity of some groups on Parler, such as QAnon and Proud Boys, on January 12, 2021, Parler was removed from the Apple and Google App Store and suspended from Amazon Cloud hosting service. Parler would have to modify their moderation policies to return to these online stores. After a month of downtime, Parler was back online with a new set of user guidelines, which reflected stricter content moderation. In this paper, we studied the moderation changes performed by Parler and their effect on the toxicity of its content. We collected a large longitudinal Parler dataset with 17M parleys from 432K active users from February 2021 to January 2022, after its return to the Internet and App Store. To the best of our knowledge, this is the first study investigating the changes in content moderation policies of Parler using data-driven approaches and also the first Parler dataset after its brief hiatus. Our quasi-experimental analysis indicates that after the change in Parler’s moderation, all forms of toxicity saw a significant decrease (𝑝 < 0.001). Finally, we found an increase in the factuality of the news sites being shared, as well as a decrease in the number of conspiracy/pseudoscience sources.

Urban fine-grained data map inference, leveraging information from coarse-grained maps, has emerged as a significant area of research due to the growing complexity and data heterogeneity in urban environments. Existing methods have a priori assumption that a coarse-grained data map, one fixed-size granularity, transforms into a fine-grained data map, also one fixed-size granularity. However, in the actual scenarios, the collected coarse-grained data maps are often incomplete and have significantly distinct granularities in various urban areas, which results in incomplete heterogeneous data, i.e., multi-granularity data maps in terms of spatial information. Meanwhile, different granularity data maps are needed for various urban downstream tasks, which is a multi-task problem. To that end, this paper proposes a novel framework, a multi-granularity super-resolution data map inference framework (MGSR), designed to harness spatio-temporal information to transform incomplete coarse-grained multi-granularity data maps into fine-grained multi-granularity data maps. Specifically, we design a granularity alignment network to align multi-granularity information and address missing data on each granularity map by leveraging the other granularity maps with a well-designed self-supervised task. Then, we introduce a feature extraction network to capture spatio-temporal dependencies and extract features. Finally, we devise a recurrent super-resolution network with shared parameters to infer multi-granularity data maps. We conduct extensive experiments on three real-world benchmark datasets and demonstrate that MGSR significantly outperforms the state-of-the-art methods for multi-granularity urban data map inference, and reduces RMSE and MAE by up to 40.1% and 50.3%, respectively. The source code has been released at https://anonymous.4open.science/r/MGSR-7E5C.

Cross-domain sequential recommendation (CDSR) is proposed to alleviate the data sparsity issue while capturing users' sequential preferences. However, most existing methods do not explore the item transition patterns across different domains and can also not be applied to a multi-domain scenario. Moreover, previous methods rely on overlapping users as bridges to transfer knowledge, which struggles to capture the complex associations across domains without sufficient overlapping users. In this paper, we introduce item attributes into CDSR, and propose a heterogeneous graph transfer learning method to address these issues. Specifically, we construct a cross-domain heterogeneous graph to allow the association of user, item, and category nodes from different domains, and enhance the flexibility of the model by enabling message propagation between more nodes through edge expansion based on the semantic similarity and co-occurrence probability. In addition, we devise meta-paths from different perspectives for nodes at item, user and category levels to guide information aggregation, which can transfer knowledge across domains and reduce the reliance on the number of overlapping users. We further design attention modules to capture users' dynamic preferences from the item sequences they have interacted with in each domain, and explore the transition patterns within category sequences which reflect users' coarse-grained preferences. Finally, we perform knowledge transfer across different domains, and predict the most likely items that users will interact with in each domain. Extensive empirical studies on three real-world datasets indicate that our HGTL significantly outperforms the state-of-the-art baselines in all cases. The source codes of our HGTL and the datasets are available at https://anonymous.4open.science/r/HGTL-C135.

WebAssembly (Wasm) is a platform-independent, low-level binary language that enables near-native performance in web applications. Given its growing importance in the web ecosystem, securing WebAssembly programs becomes increasingly important. A key security concern with WebAssembly is the threat of instruction-timing side-channel attacks, which exploit timing variations in branch instructions dependent on sensitive data, allowing attackers to infer sensitive information through timing measurement. In this paper, we introduce WaSCR, an automated WebAssembly instruction-timing Side-Channel Repairer. WaSCR uses control and data dependencies to trace the flow of sensitive data and prevent its leakage. It employs rule-based code transformations to linearize the program, eliminating branches dependent on sensitive data and substituting them with constant-time selectors. Our evaluation demonstrates that WaSCR effectively eliminates instruction-timing side channels while maintaining program correctness, with efficient repairs and moderate performance overhead.

Unsupervised graph domain adaptation (UGDA) focuses on transferring knowledge from labeled source graph to unlabeled target graph under domain discrepancies. Most existing UGDA methods are designed to adapt information from a single source domain, which cannot effectively exploit the complementary knowledge from multiple source domains. Furthermore, their assumptions that the labeled source graphs are accessible throughout the training procedure might not be practical due to privacy, regulation, and storage concerns. In this paper, we investigate multi-source-free unsupervised graph domain adaptation, i.e., exploring knowledge adaptation from multiple source domains to the unlabeled target domain without utilizing labeled source graphs but relying solely on source pre-trained models. Unlike previous multi-source domain adaptation approaches that aggregate predictions at model level, we introduce a novel model named GraphATA which conducts adaptation at node granularity. Specifically, we parameterize each node with its own graph convolutional matrix by automatically aggregating weight matrices from multiple source models according to its local context, thus realizing dynamic adaptation over graph structured data. We also demonstrate the capability of GraphATA to generalize to both model-centric and layer-centric methods. Comprehensive experiments on various public datasets show that our GraphATA can consistently surpass recent state-of-the-art baselines with different gains. Our source codes and datasets are available at https://anonymous.4open.science/r/GraphATA-C0D8.

In this paper, given a user’s query set and budget, we aim to use the limited budget to help users assemble a set of datasets that can enrich a base dataset by introducing the maximum number of distinct tuples (i.e., maximizing distinctiveness). We prove this problem to be NP-hard. A greedy algorithm using exact distinctiveness computation attains an approximation ratio of (1-e^{-1})/2, but it lacks efficiency and scalability due to its frequent computation of the exact distinctiveness marginal gain of any candidate dataset for selection. This requires scanning through every tuple in candidate datasets and thus is unaffordable in practice. To overcome this limitation, we propose an efficient machine learning (ML)-based method for estimating the distinctiveness marginal gain of any candidate dataset. This effectively eliminates the need to test each tuple individually. Estimating the distinctiveness marginal gain of a dataset involves estimating the number of distinct tuples in the tuple sets returned by each query in a query set across multiple datasets. This can be viewed as the cardinality estimation for a query set on a set of datasets, and the proposed method is the first to tackle this cardinality estimation problem. This is a significant advancement over prior methods that were limited to single-query cardinality estimation on a single dataset and struggled with identifying overlaps among tuple sets returned by each query in a query set across multiple datasets. Extensive experiments using five realworld data pools demonstrate that our algorithm, which utilizes ML-based distinctiveness estimation, outperforms all relevant baselines in effectiveness, efficiency, and scalability. A case study on two downstream ML tasks also highlights its potential to find datasets with more useful tuples to enhance the performance of ML tasks.

How might reinforcement-learning based covert social influence operations (CSIOs) be run, given that the CSIO agent wants to maximize influence and minimize discoverability of malicious accounts? And how successful can they be, given that both social platform bot detectors and humans might report them to the social platform? To answer these questions, we propose RL_CSIO, an RL-based methodology for running CSIOs and run 4 CSIOs with IRB-approval over a period of 5 days using a panel of 225 human subjects. We explore 8 research questions based on the data collected. The results show that RL_CSIO agents successfully trade off influence and discoverability - but in ways that are nuanced and unexpected.

Web-based large language model (LLM) has garnered significant attention from both academia and industry due to its potential to combine the benefits of on-device computation with the accessibility and portability of Web applications. The advent of WebGPU, a modern browser API that enables Web applications to access and utilize a device's GPU, has opened up new possibilities for GPU-accelerated LLM inference within browsers. Several frameworks have been developed to support Web-based LLM inference with WebGPU. However, our experiment reveals that these frameworks exhibit inefficiencies in GPU utilization, influencing the LLM inference speed. These inefficiencies primarily arise from underutilizing the full capabilities of WebGPU, particularly in resource management and execution synchronization. To address these limitations, we present WeInfer, an efficient Web-based LLM inference framework specifically designed to unleash the power of WebGPU. WeInfer incorporates two key innovations: 1) buffer reuse strategies that reduce the overhead associated with resource preparation, optimizing the lifecycle management of WebGPU buffers, and 2) an asynchronous pipeline that decouples resource preparation from GPU execution, enabling parallelized computation and deferred result fetching to improve overall efficiency. We conduct extensive evaluations across 9 different LLMs and 5 heterogeneous devices, covering a broad spectrum of model architectures and hardware configurations. The experimental results demonstrate that WeInfer delivers substantial improvements in decoding speed, achieving up to a $3.76\times$ performance boost compared with WebLLM, the state-of-the-art Web-based LLM inference framework.

The diffusion of culture online (e.g., hashtags) is theorized to be influenced by many interacting social factors (e.g., network _and_ identity). However, most existing computational cascade models model just a single factor (e.g., network _or_ identity). This work offers a new framework for teasing apart the mechanisms underlying hashtag cascades. We curate a new dataset of 1,337 hashtags representing cultural innovation online, develop a 10-factor evaluation framework for comparing empirical and synthetic cascades, and show that a combined network+identity model performs better than a network- or identity-only counterfactual. We also explore the heterogeneity in this result: While a combined network+identity model best predicts the popularity of cascades, a network-only model has better performance in predicting cascade growth and an identity-only model in adopter composition. The network+identity model most strongly outperforms the counterfactuals among hashtags used for expressing racial or regional identity and talking about sports or news. In fact, we are able to predict what combination of network and/or identity best models each hashtag and use this to further improve performance. In sum, our results imply the utility of multi-factor models in predicting cascades, in order to account for the varied ways in which network, identity, and other social factors play a role in the diffusion of hashtags on Twitter.

Network alignment, which aims to find node correspondence across different networks, is the cornerstone of various downstream multi-network and Web mining tasks. Most of the embedding-based methods indirectly model cross-network node relationships by contrasting positive and negative node pairs sampled from hand-crafted strategies, which are vulnerable to graph noises and leads to potential misalignment of nodes. Another line of works based on the optimal transport (OT) theory directly model cross-network node relationships and generate noise-reduced alignments. However, OT methods heavily rely on fixed, pre-defined cost functions that prohibit end-to-end training and are hard to generalize. In this paper, we aim to unify the embedding and OT-based methods in a mutually beneficial manner and propose a joint optimal transport and embedding framework for network alignment named JOENA. For one thing (OT for embedding), through a simple yet effective transformation, the noise-reduced OT mapping serves as an adaptive sampling strategy directly modeling all cross-network node pairs for robust embedding learning. For another (embedding for OT), on top of the learned node embeddings, the OT cost can be gradually trained along the learning process in an end-to-end fashion, which further enhances the alignment quality. With a unified objective, the mutual benefits of both methods can be achieved by an alternating optimization schema with guaranteed convergence. Extensive experiments on real-world networks validate the effectiveness and scalability of JOENA, achieving up to 16% improvement in MRR and 20 times speedup compared with the state-of-the-art alignment methods.

Incomplete scenario is a prevalent, practical, yet challenging setting in Multimodal Recommendations (MMRec), where some item modalities are missing due to various factors. Recently, a few efforts have sought to improve the recommendation accuracy by exploring generic structures from incomplete data. However, two significant gaps persist: 1) the difficulty in accurately generating missing data due to the limited ability to capture modality distributions; and 2) the critical but overlooked visibility bias, where items with missing modalities are more likely to be disregarded due to the prioritization of items' multimodal data over user preference alignment. This bias raises serious concerns about the fair treatment of items. To bridge these two gaps, we propose a novel Modality-Diffused Counterfactual (MoDiCF) framework for incomplete multimodal recommendations. MoDiCF features two key modules: a novel modality-diffused data completion module and a new counterfactual multimodal recommendation module. The former, equipped with a particularly designed multimodal generative framework, accurately generates and iteratively refines missing data from learned modality-specific distribution spaces. The latter, grounded in the causal perspective, effectively mitigates the negative causal effects of visibility bias and thus assures fairness in recommendations. Both modules work collaboratively to address the two aforementioneds significant gaps for generating more accurate and fair results. Extensive experiments on three real-world datasets demonstrate the superior performance of MoDiCF in terms of both recommendation accuracy and fairness. The code and processed datasets are released at https://anonymous.4open.science/r/MoDiCF-EEF5.

Information leakage issues in machine learning-based Web applications have attracted increasing attention. While the risk of data privacy leakage has been rigorously analyzed, the theory of model function leakage, known as Model Extraction Attacks (MEAs), has not been well studied. In this paper, we are the first to understand MEAs theoretically from an attack-agnostic perspective and to propose analytical metrics for evaluating model extraction risks. By using the Neural Tangent Kernel (NTK) theory, we formulate the linearized MEA as a regularized kernel classification problem and then derive the fidelity gap and generalization error bounds of the attack performance. Based on these theoretical analyses, we propose a new theoretical metric called Model Recovery Complexity (MRC), which measures the distance of weight changes between the victim and surrogate models to quantify risk. Additionally, we find that victim model accuracy, which shows a strong positive correlation with model extraction risk, can serve as an empirical metric. By integrating these two metrics, we propose a framework, namely Model Extraction Risk Inspector (MER-Inspector), to compare the extraction risks of models under different model architectures by utilizing relative metric values. We conduct extensive experiments on 16 model architectures and 5 datasets. The experimental results demonstrate that the proposed metrics have a high correlation with model extraction risks, and MER-Inspector can accurately compare the extraction risks of any two models with up to 89.58\%.

We consider the problem of bidding in online advertising, where an advertiser aims to maximize value while adhering to budget and Return-on-Spend (RoS) constraints. Unlike prior work that assumes knowledge of the value generated by winning each impression ({e.g.,} conversions), we address the more realistic setting where the advertiser must simultaneously learn the optimal bidding strategy and the value of each impression opportunity. This introduces a challenging exploration-exploitation dilemma: the advertiser must balance exploring different bids to estimate impression values with exploiting current knowledge to bid effectively. To address this, we propose a novel Upper Confidence Bound (UCB)-style algorithm that carefully manages this trade-off. Via a rigorous theoretical analysis, we prove that our algorithm achieves $\tilde{O}(\sqrt{T\log(|\mathcal{B}|T)})$ regret and constraint violation, where $T$ is the number of bidding rounds and $\mathcal{B}$ is the domain of possible bids. This establishes the first optimal regret and constraint violation bounds for bidding in the online setting with unknown impression values. Moreover, our algorithm is computationally efficient and simple to implement. We validate our theoretical findings through experiments on synthetic data, demonstrating that our algorithm exhibits strong empirical performance compared to existing approaches.

There are over a million open DNS servers in the wild. However, not all servers perform recursive queries directly. Instead, many DNS forwarders forward queries to upstream recursive servers or other DNS forwarders for name resolving on their behalf. The groups of open servers that have such dependencies on each other form ODNS Clusters. The dependencies can result in vulnerabilities; yet we have little knowledge of the ODNS cluster structure. In this work, we measure the inter-dependence of open DNS resolvers and find that 1.9 million open DNS servers form only 81,636 ODNS clusters. We further analyze the characteristics of the clustered ODNS structure. The key observations include biased cluster size distribution, discrepancy of ODNS infrastructures among countries, concentration in major public DNS server providers, and potential security and resilience risks due to the dependence.

Informal caregivers (e.g.,family members or friends) of people living with Alzheimer’s Disease and Related Dementias (ADRD) face substantial challenges and often seek informational or emotional support through online communities. Understanding the factors that drive engagement within these platforms is crucial, as it can enhance their long-term value for caregivers by ensuring that these communities effectively meet their needs. This study investigated the user interaction dynamics within two large, popular ADRD communities, TalkingPoint and ALZConnected, focusing on topic initiator engagement, initial post content, and the linguistic patterns of comments at the thread level. Using analytical methods such as propensity score matching, topic modeling, and predictive modeling, we found that active topic initiator engagement drives higher comment volumes, and reciprocal replies from topic initiators encourage further commentor engagement at the community level. Practical caregiving topics prompt more re-engagement of topic initiators, while emotional support topics attract more comments from other commentors. Additionally, the linguistic complexity and emotional tone of a comment influence its likelihood of receiving replies from topic initiators. These findings highlight the importance of fostering active and reciprocal engagement and providing effective strategies to enhance sustainability in ADRD caregiving and broader health-related online communities.

Recommender systems relying on contextual multi-armed bandits continuously improve relevant item recommendations by taking into account the contextual information. The objective of bandit algorithms is to learn the best arm (e.g., best item to recommend) for each user and thus maximize the cumulative rewards from user engagement with the recommendations. The context that these algorithms typically consider are the user and item attributes. However, in the context of social networks where $\textit{the action of one user can influence the actions and rewards of other users,}$ neighbors' actions are also a very important context, as they can have not only predictive power but also can impact future rewards through spillover. Moreover, influence susceptibility can vary for different people based on their preferences and the closeness of ties to other users which leads to heterogeneity in the spillover effects. Here, we present a framework that allows contextual multi-armed bandits to account for such heterogeneous spillovers when choosing the best arm for each user. Our experiments on several semi-synthetic and real-world datasets show that our framework leads to significantly higher rewards than existing state-of-the-art solutions that ignore the network information and potential spillover.

As privacy concerns in recommender systems become increasingly prominent, federated recommender systems (FedRecs) have emerged as a promising distributed training paradigm. FedRecs enable the collaborative training of a shared global recommendation model without requiring the exchange of raw client interaction data. However, models trained using standard FedRec methods remain vulnerable to personal information leakage, particularly through attribute inference attacks, which can expose sensitive user attributes such as gender and race. In this paper, we address these user sensitive attributes as targets for federated unlearning. To protect users' sensitive information, attribute unlearning aims to eliminate sensitive attributes from user embeddings, thereby preventing inference attacks while preserving recommendation performance. We introduce a novel post-training federated unlearning framework, Aegis, which performs unlearning based on private attribute requests after the model has been trained, minimizing the degradation in recommendation accuracy. Aegis employs an information-theoretic multi-component loss function to balance privacy protection and recommendation performance. Additionally, Aegis adapts to scenarios where training interaction data may be unavailable, reflecting real-world centralized protection scenarios. Comprehensive evaluations on various benchmark datasets demonstrate that our proposed method effectively safeguards user privacy while maintaining high-quality recommendations.