Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SpamDam: Towards Privacy-Preserving and Adversary-Resistant SMS Spam Detection

Yekai Li, Rufan Zhang, Wenxin Rong, Xianghang Mi

TL;DR

SpamDam tackles the dual challenges of privacy and adversarial risk in SMS spam detection by building an end-to-end framework with four integrated modules: a cross-OSN spam radar, a statistical inspector, multilingual detectors including binary and multi-label classifiers, and an adversarial analyzer. It delivers the largest up-to-date SMS spam dataset (76,577 messages from Twitter and Weibo across 2018–2023), demonstrates strong multilingual detection performance, and proves the practicality of federated learning for privacy-preserving training. The study also rigorously assesses adversarial robustness, introducing a reverse backdoor attack and showing that adversarial training and data sanitization are essential defenses. Overall, SpamDam advances privacy-preserving, adversary-resistant SMS spam detection with a scalable data resource, multilingual models, and security-oriented evaluation that informs deployment in privacy-sensitive and crowdsourced-data settings.

Abstract

In this study, we introduce SpamDam, a SMS spam detection framework designed to overcome key challenges in detecting and understanding SMS spam, such as the lack of public SMS spam datasets, increasing privacy concerns of collecting SMS data, and the need for adversary-resistant detection models. SpamDam comprises four innovative modules: an SMS spam radar that identifies spam messages from online social networks(OSNs); an SMS spam inspector for statistical analysis; SMS spam detectors(SSDs) that enable both central training and federated learning; and an SSD analyzer that evaluates model resistance against adversaries in realistic scenarios. Leveraging SpamDam, we have compiled over 76K SMS spam messages from Twitter and Weibo between 2018 and 2023, forming the largest dataset of its kind. This dataset has enabled new insights into recent spam campaigns and the training of high-performing binary and multi-label classifiers for spam detection. Furthermore, effectiveness of federated learning has been well demonstrated to enable privacy-preserving SMS spam detection. Additionally, we have rigorously tested the adversarial robustness of SMS spam detection models, introducing the novel reverse backdoor attack, which has shown effectiveness and stealthiness in practical tests.

SpamDam: Towards Privacy-Preserving and Adversary-Resistant SMS Spam Detection

TL;DR

SpamDam tackles the dual challenges of privacy and adversarial risk in SMS spam detection by building an end-to-end framework with four integrated modules: a cross-OSN spam radar, a statistical inspector, multilingual detectors including binary and multi-label classifiers, and an adversarial analyzer. It delivers the largest up-to-date SMS spam dataset (76,577 messages from Twitter and Weibo across 2018–2023), demonstrates strong multilingual detection performance, and proves the practicality of federated learning for privacy-preserving training. The study also rigorously assesses adversarial robustness, introducing a reverse backdoor attack and showing that adversarial training and data sanitization are essential defenses. Overall, SpamDam advances privacy-preserving, adversary-resistant SMS spam detection with a scalable data resource, multilingual models, and security-oriented evaluation that informs deployment in privacy-sensitive and crowdsourced-data settings.

Abstract

In this study, we introduce SpamDam, a SMS spam detection framework designed to overcome key challenges in detecting and understanding SMS spam, such as the lack of public SMS spam datasets, increasing privacy concerns of collecting SMS data, and the need for adversary-resistant detection models. SpamDam comprises four innovative modules: an SMS spam radar that identifies spam messages from online social networks(OSNs); an SMS spam inspector for statistical analysis; SMS spam detectors(SSDs) that enable both central training and federated learning; and an SSD analyzer that evaluates model resistance against adversaries in realistic scenarios. Leveraging SpamDam, we have compiled over 76K SMS spam messages from Twitter and Weibo between 2018 and 2023, forming the largest dataset of its kind. This dataset has enabled new insights into recent spam campaigns and the training of high-performing binary and multi-label classifiers for spam detection. Furthermore, effectiveness of federated learning has been well demonstrated to enable privacy-preserving SMS spam detection. Additionally, we have rigorously tested the adversarial robustness of SMS spam detection models, introducing the novel reverse backdoor attack, which has shown effectiveness and stealthiness in practical tests.
Paper Structure (16 sections, 4 figures, 13 tables)

This paper contains 16 sections, 4 figures, 13 tables.

Table of Contents

  1. Introduction
  2. Background
  3. Collecting SMS Spam
  4. Privacy-Preserving SMS Spam Discovery
  5. Understanding Up-To-Date SMS Spam
  6. Multilingual SMS Spam Classification
  7. Binary SMS Spam Classification
  8. Multi-Label SMS Spam Classification
  9. SMS Spam Detection via Federated Learning
  10. Adversary-Resistant SMS Spam Classification
  11. Adversarial Examples
  12. Poisoning Attacks
  13. Privacy-Preserving SMS Spam Classification
  14. Discussion
  15. Related Work
  16. ...and 1 more sections

Figures (4)

  • Figure 1: SpamDam: a framework to enable privacy-preserving and adversary-resistant SMS spam detection.
  • Figure 2: The temporal evolution of SMS spam messages as observed on both Twitter and Weibo. For 2023 H2, we only have data for the first three months (July to September).
  • Figure 3: The effectiveness of adversarial examples and adversarial training on SMS spam detection.
  • Figure 4: The effectiveness of reverse backdoor attacks via injecting stamped benign messages under the label of spam.