This paper addresses the challenge of reliability in open rating systems on digital platforms, which are susceptible to low-quality and malicious ratings. We introduce a novel trust-aware filtering mechanism that incorporates both user rating behavior and trust networks into a directed weighted graph model. The core contribution of our approach lies in the integration of “anchor users,” who serve as stable points of trust beyond dynamic online interactions. We further enhance our methodology by introducing a clustering algorithm specifically designed for this application scenario, and a novel method of filtering based on clustering. This involves examining intra-cluster ratings to achieve more fine-grained and filtered rating aggregates, effectively filtering out adversarial ratings and enhancing the reliability of the rating system. Unlike traditional systems that rely on a static hierarchy of trust, our method uses a personalized set of anchor users for each individual, offering a scalable solution to the diversity of trust relationships in growing user bases. Our contributions are demonstrated through an experimental setup that validates the effectiveness of our approach in improving the quality and personalization of content in open rating systems. The paper further discusses the application scenarios and concludes with the potential for further exploration in enhancing open rating system reliability.