Blockchain Based Privacy Preserving Recommendation System (BPPRS) Using Digital Signatures

Abstract

Recommendation system plays an important role in the digital world, since it helps to find users interest. Collaborative filtering is the widely used algorithms in recommendation systems due to its simplicity and efficiency. However, when the user rating data is sparse, which leads to generate unreasonable recommendations for those users who provide no ratings. Faced with these problems, the proposed Blockchain based Privacy Preserving Recommendation System (BPPRS) model is built using a new collaborative filtering algorithm for recommendation by combining the Jaccard Similarity and Triangle Similarity (JSTS). Using the values calculated from JSTS, an item similarity matrix is constructed in order to find the zero rated users so that the item name of these users are recommended to the particular user who are closely similar. The experimental result shows that the proposed model mitigates the sparseness of the information. Further, the BPPRS model is enabled by the use of Blockchain technology to generate the hash value for each of the items recommended to the user; previous hash value is used for current recommended items hash value and stored in different blocks. The encryption of the data ensures the secure recommendation and also the use of MECDSA makes it a better trap door function.