This thesis proposal aims to develop a new similarity computation method for collaborative recommender systems that incorporates the use of trust between users. The proposed method will have three phases: 1) constructing neighborhoods of similar users using Pearson correlation and Jaccard similarity, 2) determining trust values for neighborhood members based on similarity, confidence, and profile trust measurements, and 3) combining phase 1 and 2 to define trusted neighbors and calculate actual similarities. The goal is to improve system accuracy by leveraging trust relationships. A literature review discusses existing trust-based recommendation approaches.