Title :
Forensic outlier detection for Cognitive Radio Networks
Author :
Kabir, I. ; Astaneh, S. ; Gazor, S.
Author_Institution :
Dept. of Electr. & Comput. Eng., Queen´s Univ., Kingston, ON, Canada
Abstract :
We consider forensic outlier detection instead of traditional outlier detection to enforce spectrum security in a Cognitive Radio Network (CRN). We investigate a CRN where a group of sensors report their local binary decisions to a Fusion Center (FC), which makes a global decision on the availability of the spectrum. To ensure the truthfulness of the sensors, we examine the reported decisions in order to determine whether a specific sensor is an outlier. We propose several optimal detectors (for known parameters) and suboptimal detectors (for the practical cases where the parameters are unknown) to detect three types of outlier sensors: 1) selfish sensor, which reports the spectrum to be occupied when locally detects its vacancy, 2) malicious sensor, which reports the spectrum to be vacant when locally detects its occupancy, 3) malfunctioning sensor, whose reports are not accurate enough (i.e., its performance is close to random guessing). We evaluate the proposed detectors by simulations. Our simulation results reveal that the proposed detectors significantly outperform the Grubb´s test. Since the unknown or untrustworthy parameters are accurately estimated by the FC, the proposed suboptimal detectors do not require the knowledge of the spectrum statistics and are insensitive to the parameters reported by the suspected user. These detectors can be used by government agencies for forensic testing in policy control and abuser identification in CRNs.
Keywords :
cognitive radio; decision theory; radio networks; sensor fusion; signal detection; telecommunication security; CRN; FC; Grubb test; abuser identification; cognitive radio networks; forensic outlier detection; forensic testing; fusion center; local binary decisions; malfunctioning sensor; malicious sensor; optimal detectors; outlier sensors; policy control; selfish sensor; spectrum security; spectrum statistics; suboptimal detectors; Availability; Cognitive radio; Detectors; Forensics; Maximum likelihood estimation; Simulation; Cognitive radio; forensic cognitive detection; outlier detection; policy enforcement; spectrum security;
Conference_Titel :
Communications (QBSC), 2014 27th Biennial Symposium on
Conference_Location :
Kingston, ON
DOI :
10.1109/QBSC.2014.6841183