Abstract: China Space Survey Telescope (CSST) is a 2-meter aperture optical space telescope planned by the China Manned Space Program, with the primary focus on large-scale surveys, which combines high spatial resolution and a large field of view. Transients are one of the important scientific objectives of CSST. As a space telescope, the image noise composition of CSST is significantly different from that of ground-based telescopes. Therefore, it is unsuitable to directly apply the transient detection scheme of previous ground-based survey project, and it is necessary to develop an independent detection method for CSST. The CSST transient detection method is developed and validated based on the CSST simulation data, aiming to identify transients in CSST main survey data as accurately as possible. The method is based on Difference Image Analysis (DIA) method, which firstly obtains the residual images corresponding to the observed image, and then uses filtering method to obtain the local Poisson noise distribution of the residual image. The method then uses a changing threshold to indirectly unify the image noise level, and performs photometry and source extraction on the residual image to obtain a list of candidate sources. Next, the real sources are filtered by judging the deviation of the candidate sources from local Poisson noise. After combining the photometric classification information of the observed image and the information of original image, the fake sources in the candidate source list are further eliminated. The final output is a transient list. The method is able to filter transients in the observed images with an average accuracy of 95.9% in a multi-round simulation test with a total of 20000 transients. Compared with the previous work, the test also quantifies the detection rates of transients with different brightness more comprehensively. The conclusions of the test validate the feasibility, generalisation and stability of the CSST transient detection method. The method also provides a theoretical and programming basis for transient detection mission of space telescopes.