The invention relates to a ciphertext sorting searching scheme based on a classification document in a cloud network. According to the scheme, through a document clustering method, after a document isclassified, key words are extracted to establish an index, an original high-dimension secret key is converted into a low-dimension secret key, and the time of establishing the index is optimized. According to the scheme, by adopting a scoring mechanism in a key word group, the flexibility of adding and deleting the document is improved. During index updating, a group vector corresponding to a type key word set which needs to be changed is updated, and the time of updating the index is shortened. In the retrieval aspect, a targeted retrieval scheme is provided, a cloud server does not need tocalculate and query the whole vector volume of a trap door and the index during searching but only needs to calculate the volume of the group vector containing the query key word, the retrieval speedis increased, and the retrieval efficiency is improved.