Tampering with the data of electricity meters is a typical electricity theft. For such electricity theft behaviors, the existing detection methods require a labeled data set or additional power system state information, which is difficult to obtain or has a large error with the actual value. Therefore, it is urgent to utilize the lower dimension data to realize the detection of electricity theft behavior. In this paper, a novel fusion detection method is proposed by combining the maximum information coefficient (MIC) technique and the clustering by fast search and find of density peaks. This method uses MIC to measure the correlation between management line loss and specific behaviors of consumers and uses CFSFDP to locate abnormal electricity consumers with high applicability, which can detect various types of electricity theft. This paper also uses the Irish smart meter data set to verify the algorithm, and the good performance of the proposed method is proved by the result.