The data mining system can help to identify the important hidden patterns for decision-making in large datasets. Privacy preserving data mining (PDDM) has emerged as a critical area for the sharing, decision-making and dissemination of confidential data. Preserving privacy is a common data security model for protecting unauthorised access to individual decision patterns. Because the distributed data of the individuals is processed by the third Party, the information in digital networks is misused. Such information on privacy about businesses , industries and persons must be encoded prior to publication or published. During the processing of data from various sources, decision patterns based on standard data security protection models such as Naïve Bayes, SVM and the models for the decision tree are very difficult to maintain. In addition, the use of traditional models to fill sparse values is inefficient and inadequate for the protection of privacy. A novel data security model was developed and applied on large data sets in this paper. In this model a philtre based data protection scheme is used to cover decision patterns with homomorphic encoding and decryption algorithm using the decision-tabo classifier. In this scheme. Experimental findings demonstrated the high processing efficiency of the proposed model relative to conventional data protection approaches of large-scale datasets.