The widespread use of communication systems increases the usage of the frequency spectrum. Measurements reveal that the usage of the spectrum is av eragely less than 15% and in the the maximum case the usage is close to 85%. Hence,the main problem in spectrum allocation is not the shortage of the frequency spectrum but it is the inefficient usage. This is because by now fixed frequency spectrum has been assigned to the licensed users or primary users. Three Proposed solutions to this problem are spectrum reallocation, spectrum leasing and spectrum sharing. Efforts for more efficient usage of spectrum through spectrum sharing necessiate‌‌the use of cognitive radio networks. Cognitive radio is an advanced radio system that can adjust its operating parameters for clever access to the spectrum. One of the essential functions ofeach cognitive radio network is spectrum sensing to detect the presence or absence of the primary user signal. Cooperative spectrum sensing is used to avoid the effects of the channel fading in spectrum sensing.. In this method, the reports from several cognitive radio users are collected and the final decision about the presence or absence of the primary user signal is made.In this waymalicious users cause mistake on the desicion made on the presence or absence of theprimary user signal. Therefore methods have been presented in cooperative spectrum sensing to eliminate the effect of the attacks by malicious users.As proposed methods for eleminatingthe effect of malicious users are outlier tests, reputation based methods and methods provided for mobile networks. In the outlier tests, identification of malicious users while they are the majority of users is hardly possible. Nevertheless, the simplicity and the rapid detection of the malicious users are the advantages of outlier tests. Outlier tests which have distance approach have a good performance for identification of the malicious users in cognitive radio networks. Local outlier factor test (LOF) is as one of outlier tests with distance approach.In comparison with other outlier tests such as ABP and GESD tests, local outlier factor test (LOF) has better performance in detecting malicious users based on the curves of the detection probability vs. false alarm probability. In this thesis GLOF test that is the generalization of LOF test is used to detect malicious users.It is shown that when the number of the malicious users are high GLOF test has better performance compared with LOF test. Furthermore, PLOF test presented in this thesis uses peeling method with LOF test. In this test, the most outlier data at every step is detected and will be deleted from data set ofthe cognitive radio users reports. Such steps are repeated until all the outlier data are removed. It is shown that when the number of the malicious users is less, PLOF method is better than LOF and GLOF. LOF and GLOF tests have the same computational complexity; however, PLOF test has higher computational complexity.The proposed methods provide detection of malicious users properly in a cognitive radio network with simplicity and rapid spectrum sensing. Keywords: Cognitive radio, cooperative spectrum sensing, outlier tests, onion peeling, generalized local outlier factor