Abstract: While the advanced sensing and communication functions improve the measurement and control functions, the distribution network is also faced with various possible network attacks. When the attacker tampers with other users and their own measurement records by specific means, so as to achieve the electricity stealing behavior under the condition of keeping the total electricity data unchanged, the traditional electricity stealing monitoring method based on power balance is difficult to effectively identify. Therefore, on the basis of the high-order statistical information of smart meter measurement data, this paper proposes a non-invasive detection method of power larceny in distribution network based on smart meter measurement. It can detect the attacker and the victim. Experiments show that the proposed method can effectively detect non-invasive attacks by using higher-order statistics of power consumption, and can identify the users (attackers and victims) involved.