Sensitive Information Securit y (SIS) model provides the strong base to transmit a large volume of sensitive information based textual documents securel y and safel y . In this paper, a model is proposed to provide sensitive information securit y using new techniques for dimensionalit y reduction where two new algorithms Term Similarit y Clustering (TSC) and Term Index Clustering (TIC) based on supervised learning are proposed for SIS model. The TSC approach includes the loss y data reduction at the sender side, but makes the s y stem tolerable to maintain integrit y b y keeping confidentialit y at its best level when data is sent over on the unsecured communication channel. The terms or words are extracted from the text documents and the y are categorized into Term Clusters (TC) with the use of Knowledge Repositor y (KR). Otherwise, it uses the KR and then makes a new entr y into the appropriate TC with its associated TSC. Then these TCs of reduced dimension are provided the high securit y and sent over on the unsecured channel. The second approach TIC works differentl y and provides the loss less data reduction at the sender side with high data integrit y and confidentialit y during data transmission but includes an overhead of keeping KR at the receiver side. In TIC, instead of keeping the word in the TC, the index of each term, referring the knowledge repositor y , is placed in the respective TIC. Due to the KR overhead, this approach increases the configuration complexit y at the receiver side. As both proposed approaches decrease the space and time complexities, so this paper provides the anal y tical experimental results of testing a text document of NASA Standards in which TSC gives about 10% space dimension reduction with some information loss, whereas TIC provides around 12% of space reduction with the additional overhead of KR at the receiver side.