Identifying malicious users accurately in cognitive radio networks(CRNs) is the guarantee for excellent detection performance. However, existing algorithms fail to take the mobility of secondary users into consideration. If applied directly in mobile CRNs, those conventional algorithms would overly punish reliable users at extremely bad or good locations, leading to an obvious decrease in detection performance. To overcome this problem, we divide the whole area of interest into several cells to consider the location diversity of the network. Each user’s reputation score is updated after each sensing slot and is used for identifying whether it is malicious or not. If so, it would be removed away. And then our algorithm assigns users in cells with better channel conditions, i.e. larger signal-to-noise ratios(SNRs), with larger weighting coefficients, without requiring the prior information of SNR. Detailed analysis about the validity of our algorithm is presented. The simulation results show that in a CRN with 60 mobile secondary users, among which, 18 are malicious, our solution has an improvement of detection probability by 0.97-d B and 3.57-d B when false alarm probability is 0.1, compared with a conventional trust-value-based algorithm and a trusted collaborative spectrum sensing for mobile CRNs, respectively. However, existing algorithms fail to take the mobility of secondary users into consideration. If applied directly in mobile CRNs, those conventional algorithms would overly punish reliable users at To overcome this problem, we divide the whole area of ​​interest into several cells to consider the location diversity of the network. Each user’s reputation score is updated after each sensing slot and is then used for identifying whether it is malicious or not. And then our algorithm assigns users in cells with better channel conditions, ie larger signal-to-noise ratios (SNRs), with larger weighting coefficients, without requiring the prior information of SNR. Detailed analysis about the validity of our algorithm is presented. The simulation results show that in a CRN with 60 mobile secondary users, among among, 18 are malicious, our solution has an improvement of detection probability by 0.97-d B and 3.57-d B when false alarm probability is 0.1, compared with a conventional trust-value -based algorithm and a trusted collaborative spectrum sensing for mobile CRNs, respectively.