In cooperative perception systems based on Vehicle-to-Everything (V2X) technologies, a critical task is assessing agent reliability to protect against Byzantine attacks — malicious or faulty participants transmitting false data. This paper proposes a multi-factor Trust Score — a quantitative trust metric integrating three independent consistency components: sensor consistency (comparison of V2X data with own measurements), consensus consistency (agreement with neighboring agents’ data), and physics consistency (compliance with vehicle dynamics laws). A detailed description of the system architecture, theoretical justification of each component, and analysis of applicability conditions are presented. A Byzantine-resilient data aggregation algorithm with formal correctness guarantees is developed. Special attention is given to practical implementation aspects and parameter selection.