Guest editorial: special issue on trust, privacy, and security in crowdsourcing computing

An Liu, Guanfeng Liu, Mehmet A. Orgun, Qing LiWorld Wide Web 2020 · OpenReview

Introduction

Crowdsourcing has emerged as a dominant paradigm for a wide range of digital tasks, from data labeling and named entity recognition to mobile sensing and spatial data collection. However, because these systems rely on cooperation between often anonymous individuals with varying motivations, they are inherently vulnerable to security breaches and privacy leaks. This article synthesizes the contributions of a special issue of the World Wide Web journal, which features 12 high-quality papers addressing the "trust deficit" in modern crowdsourcing environments.

The Problem: The Trust Deficit in Human-Centric Systems

While crowdsourcing offers unparalleled efficiency and scalability, its human-centric nature introduces significant risks. Requesters and workers are often strangers, leading to potential conflicts of interest. Malicious actors may attempt to deanonymize users, provide low-quality data for rewards, or engage in "crowdturfing"—coordinated efforts to spread fake reviews or manipulate social media activity. The core challenge lies in maintaining system integrity and user privacy without sacrificing the utility of the collected data.

Methodological Approaches

The special issue highlights a diverse array of technical strategies designed to secure crowdsourcing ecosystems:

  • Decentralized Trust: Several studies explore the use of Blockchain technology to establish transparent and immutable records of transactions and data quality, reducing the need for a central trusted authority.
  • Privacy-Preserving Cryptography: Researchers implemented advanced techniques such as homomorphic encryption for cloud-based graph matching and secure range queries, ensuring that sensitive data remains encrypted even during processing.
  • Differential Privacy: To protect spatial data, authors utilized "indefeasible" Laplace noise, allowing for accurate geographic analysis while preventing the identification of individual contributors.
  • Graph Analytics and Machine Learning: New distributed algorithms based on GraphX were developed for pattern matching in massive temporal graphs. Additionally, semi-supervised learning models were introduced to detect spammers by analyzing behavioral features and network properties.
  • Economic Incentives: To ensure data integrity, some papers proposed reverse-auction mechanisms that incentivize truthful labeling in active learning environments.

Key Insights and Results

The research presented in this special issue offers several critical insights into the future of secure computing:

The Privacy-Utility Trade-off

A recurring theme is the search for an "optimal trade-off." Security measures must be robust enough to protect providers but flexible enough to ensure the data remains valuable to the requester. For instance, spatial decomposition techniques allow for useful location-based services while masking exact user coordinates.

Robustness and Scalability

In the realm of knowledge graphs, where "facts noise" is a common hurdle, ensemble strategies proved more effective than standard embedding methods for maintaining data trust. Furthermore, the introduction of distributed algorithms demonstrated that security measures can scale to handle "massive temporal graphs" without significant performance degradation.

Detection and Algorithmic Security

The special issue reports high success rates in identifying coordinated crowdturfing campaigns. On the algorithmic front, the development of "fully secure" range query methods now allows for the preservation of confidentiality regarding the query itself, the results, and even the underlying access patterns.

Future Directions

Despite these advancements, the editors identify several areas requiring further exploration:

  • Practical Blockchain Integration: Implementing blockchain in high-velocity, real-time crowdsourcing environments remains a technical challenge.
  • Mobile Sensing Gaps: There is a pressing need for more practical privacy preservation mechanisms specifically tailored for mobile crowd sensing.
  • Dynamic Environments: Future research must move beyond static models to address trust and security in highly dynamic, real-time applications where user behavior and network conditions change rapidly.

By curating these 12 papers from a competitive pool of 26 submissions, this special issue provides a comprehensive snapshot of the current state of security in crowdsourcing, offering a roadmap for more resilient and trustworthy human-centric computing.

Summary updated

BibTeX
@inproceedings{liu2020guest,
  author = {An Liu and Guanfeng Liu and Mehmet A. Orgun and Qing Li},
  title = {{Guest editorial: special issue on trust, privacy, and security in crowdsourcing computing}},
  booktitle = {World Wide Web},
  year = {2020},
  url = {https://openreview.net/forum?id=UUSEAdn8Or},
}
Figure