This paper aims at presenting a novel technique to find duplicate records in hierarchical (XML) data which contains multimedia attributes. Now a days the data is being stored in more complex and semistructured or hierarchical structure and  the  problem  arose  is  how  to  detect  duplicates  on  this  XML  data.  Due to differences between various data models we cannot apply same algorithms which are for single relation on XML data. The objective  of  this  paper  is  to  detect  duplicates  in  hierarchical  data  which  contain  textual  and multimedia data like images, audio and video. Also to act according to user choice on that data like delete, update etc.  Also  to  prune  the  duplicate  data  by  using  pruning  algorithm  that  is  included  in proposed system. Here Bayesian network will be used for duplicate detection, and by experimenting on both artificial and real world datasets the MULTIDUP method will be able to perform duplicate detection with high efficiency and effectiveness. This method will compare each level of XML tree from root to the leaves. The first step is to go through the structure of tree comparing each descendant of both datasets inputted and find duplicates despite difference in data.