Catching Attacker(s) for Collaborative Spectrum Sensing in Cognitive Radio Systems: An Abnormality Detection Approach

Collaborative spectrum sensing, which collects local observations or decisions from multiple secondary users to make a decision by a fusion center, is an effective approach to alleviate the unreliability of single-user spectrum sensing. However, it is subject to the attack of malicious secondary user(s), which may send false reports. Therefore, it is necessary to detect potential attacker(s) and make attack-proof decisions for spectrum sensing. Most existing attacker detection schemes are based on the knowledge of the attacker's strategy and thus apply the Baeysian detection of attackers. However, in practical cognitive radio systems, the data fusion center typically does not know the attacker's strategy. To alleviate the problem of the unknown strategy of attacker(s), an abnormality detection approach, based on the abnormality detection in data mining, is proposed. The performance of the attacker detection in the single-attacker scenario is analyzed explicitly. For the case that the attacker does not know the reports of honest secondary users (called independent attack), it is numerically shown that attacker can always be detected as the number of spectrum sensing rounds tends to infinity. For the case that the attacker knows all the reports of other secondary users, based on which the attacker sends its report (called dependent attack), an approach for the attacker to perfectly avoid being detected is found, provided that the attacker has perfect information about the miss detection and false alarm probabilities. This motivates cognitive radio systems to protect the reports of secondary users. The performance of attacker detection in the general case of multiple attackers is demonstrated using numerical simulations.