Blind Dogfight in Spectrum: Combating Primary User Emulation Attacks in Cognitive Radio Systems with Unknown Channel Statistics

Primary user emulation (PUE) attack is a serious threat to cognitive radio systems. A passive anti-PUE approach, similar to the random frequency hopping in traditional anti-jamming schemes, is proposed. The defenders randomly choose channels to sense and avoid the PUE attack while the strategy of the attacker is arbitrary. Channel statistics like availability probabilities are assumed unknown to both the attacker and defenders. The environment of each defender, including the attacker, channels and other defenders, is unknown to the defender and is then modeled as an adversarial multi-armed bandit. The technique of adversarial bandit problem is applied in the defenders' strategies to alleviate the problem of unknown environment. Lower bounds of performance for defenders, subject to several typical attack strategies, are derived for single defender case. Numerical simulations demonstrated the performance of the proposed attack avoidance scheme for both the single defender and multiple defender cases.