Multiple identification and attenuation are one of the most challenging tasks in the seismic data processing chain. We are presenting an approach for the identification of surface related multiples within the Common Reflection Surface (CRS) workflow, so that a processing chain from time to depth imaging with CRS related technology can be established. The approach is based on multiple prediction by auto-convolving each stacked trace. The implementation of the method is done with focus on automating the process. After the multiples are identified/predicted their offset traveltimes are calculated with the help of the CRS attributes, afterwards adaptively filtered, and subtracted from the prestack data. Initial tests were performed on a real marine data set. The results indicate the potential of further developing this method. The identified/predicted multiples could be successfully removed from the data set. The prediction by auto-convolution is simple and robust, but is also affected by some prediction errors due to theoretical assumptions involved in this concept.