<p>This study introduces a privacy-preserving recommender system solution designed to be both effective and capable of ensuring the privacy of honest users in the face of potentially malicious behaviour by the server. At the stage of building the recommender system, we propose a protocol that allows calculating multiple average ratings and item-item similarity values with privacy-preserving in one execution. In the proposed protocol, we perform optimization in the use of private and public key pairs of the participating parties. This will reduce the computational cost and communication costs of the process of building a recommender system. In addition, our solution proposes to use the elliptic curve cryptosystem which will significantly reduce the communication cost. Notably, the proposed solution maintains privacy even when the server collaborates with up to n-2 corrupted users, protecting non-colluding parties. Theoretical and experimental analyses demonstrate that this solution not only preserves the accuracy of the recommender system but also significantly enhances its overall performance. This solution may be used in a variety of areas, including data science, e-commerce, medicine/health care, finance, decision support systems, and supply chain management.</p>