Techniques for generating a data representation without access to content are described. A method for generating a data representation without access to content comprises receiving a request to analyze one or more data items in a protected area of the provider network, sending the request to the protected area of the provider network, wherein the cluster model is used to identify a cluster identifier associated with each of the one or more data items, receiving the cluster identifier associated with each of the one or more data items, and regenerating each of the one or more data items based on the cluster identifier.