Cooperative spectrum sensing is vulnerable to attacks from malicious nodes, especially when collusion occurs. In this paper, we analyze the effect of colluded statistical attacks and show that collusion could cause performance degradation in terms of both false-alarm and detection probabilities, which is not possible via independent attacks. Closed-form expressions for system performance under the majority fusion rule are provided for a generalized form of colluded attacks. Then, for specific scenarios of collusion and mimicry attacks, we study the conditions under which the probabilities of false alarm and detection are both degraded.