Robust spectrum sensing with crowd sensors

Guoru Ding, Jinlong Wang, Qihui Wu, Linyuan Zhang, Yulong Zou, Yu Dong Yao, Yingying Chen

Research output: Contribution to journalArticlepeer-review

128 Scopus citations

Abstract

This paper investigates the issue of cooperative spectrum sensing with a crowd of low-end personal spectrum sensors (such as smartphones, tablets, and in-vehicle sensors), where the sensing data from crowd sensors that may be unreliable, untrustworthy, or even malicious. Moreover, due to either unexpected equipment failures or malicious behaviors, every crowd sensor could sporadically and randomly contribute with abnormal data, which makes the existing cooperative sensing schemes ineffective. To tackle these challenges, we first propose a generalized modeling approach for sensing data with an arbitrary abnormal component. Under this model, we then analyze the impact of general abnormal data on the performance of the cooperative sensing, by deriving closed-form expressions of the probabilities of global false alarm and global detection. To improve sensing data quality and enhance cooperative sensing performance, we further formulate an optimization problem as stable principal component pursuit, and develop a data cleansing-based robust spectrum sensing algorithm to solve it, where the under-utilization of licensed spectrum bands and the sparsity of nonzero abnormal data are jointly exploited to robustly cleanse out the potential nonzero abnormal data component from the original corrupted sensing data. Extensive simulation results demonstrate that the proposed robust sensing scheme performs well under various abnormal data parameter configurations.

Original languageEnglish
Article number6876192
Pages (from-to)3129-3143
Number of pages15
JournalIEEE Transactions on Communications
Volume62
Issue number9
DOIs
StatePublished - 1 Sep 2014

Keywords

  • Cognitive radio networks
  • data cleansing
  • data quality
  • mobile crowd sensing
  • spectrum sensing

Fingerprint

Dive into the research topics of 'Robust spectrum sensing with crowd sensors'. Together they form a unique fingerprint.

Cite this