In Cognitive radio (CR) networks, Byzantine attack is a threat to the spectrum sensing. In the so-called attack, some malicious CR users intentionally report wrong local spectrum sensing results to a fusion center (FC) and considerably disturb the global spectrum sensing decision. In this paper, we investigate three different types of Byzantine attacks, always yes (AY), always no (AN) and always false (AF) attackers. The AY attackers increase the global false alarm probability and the AN malicious attackers decrease the global detection probability. In addition, the AF Byzantine attackers decrease the global detection probability and increase the global false alarm rate. We carry out simulations with a conventional majority fusion rule under three types of Byzantine attack strategies. We also perform numerical analyses to illustrate these attack scenarios on the global detection, false alarm and correct sensing probabilities.