EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

ARGUS: A Robust Guard for Mobile Crowdsensing User Registration

Cihan Eryonucu, Panos Papadimitratos

IEEE Access · 2026 · 14 · 110856-110878

DOI: 10.1109/access.2026.3714741 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Cihan Eryonucu und 1 weiteren Autor:innen. Die Arbeit nennt 95 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0). Gefördert von: Wallenberg AI, Autonomous Systems and Software Program, Knut och Alice Wallenbergs Stiftelse.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Mobile Crowdsensing Systems (MCS) can be attacked by an adversary that registers numerous fictitious, typically emulated, devices to report bogus data in large volumes. It is crucial to safeguard MCS by preventing registration of fictitious devices, broadly termed Sybils. Most existing defenses try to detect Sybils essentially too late, after the adversary succeeds in having them join and after having influenced the system; more so, they require sensitive user information or demand burdensome manual tasks, ultimately discouraging participation. This is why we introduce ARGUS as an early screening mechanism that thwarts attempts to register Sybils at a high rate at the first stage of user registration. ARGUS reduces the fictitious device registration rate by filtering out their registrations early in their interaction with the system, before their adversarial data can influence the MCS. To sift fake/Sybil users, we utilize sparsely sampled smartphone battery logs to reflect actual user and smartphone usage characteristics without exposing privacy-sensitive information. ARGUS isolates and identifies even (colluding) Sybil users orchestrated by an attacker with a high volume of synthetically generated data. ARGUS leverages well-known machine learning components in a novel way and provides protection for systems without relying on historical user data or user interaction (as required, notably, by CAPTCHA); that is, without reducing privacy, inconveniencing users, or discouraging participation. We evaluate the effectiveness of ARGUS through extensive experiments, demonstrating it consistently achieves an F1-score greater than 90% in detecting a broad range of Sybil registrations; with adversarial strategies ranging from perturbing genuine traces to training generative adversarial networks (GANs). Unlike existing defenses that verify device integrity (e.g., device attestation) or human presence (e.g., CAPTCHA), both of which are susceptible to high bypass rates under certain attacks, ARGUS assesses behavioral authenticity through battery usage patterns, without user interaction and pre-emptively, offering a new, fundamentally orthogonal, and complementary countermeasure. Lightweight and easy to deploy, ARGUS is the first method to leverage mobile device battery traces to detect Sybil attacks in MCS systems without exposing privacy-sensitive user data. ARGUS can significantly increase the cost for any adversary that seeks to introduce Sybils at a high rate, effectively and efficiently detecting and rejecting those that result from adversarial automation and generated data.

Förderung

  • Wallenberg AI, Autonomous Systems and Software Program
  • Knut och Alice Wallenbergs Stiftelse

Zitieren

APA

Eryonucu, C., & Papadimitratos, P. (2026). ARGUS: A Robust Guard for Mobile Crowdsensing User Registration. IEEE Access, 14, 110856-110878. https://doi.org/10.1109/access.2026.3714741

Vancouver

Eryonucu C, Papadimitratos P. ARGUS: A Robust Guard for Mobile Crowdsensing User Registration. IEEE Access. 2026;14:110856-110878. doi:10.1109/access.2026.3714741

BibTeX
@article{Eryonucu2026,
  author = {Eryonucu, Cihan and Papadimitratos, Panos},
  title = {ARGUS: A Robust Guard for Mobile Crowdsensing User Registration},
  journal = {IEEE Access},
  year = {2026},
  volume = {14},
  pages = {110856--110878},
  doi = {10.1109/access.2026.3714741}
}
Kostenlos starten →