In urban rail transit systems, Communication-Based Train Control (CBTC) utilizes bidirectional wireless communication to enable real-time data exchange between trains and trackside equipment, thereby ensuring normal train operations. However, the open operating environment of CBTC exposes it to the threat of cyber attacks. To address this issue, this paper proposes a collaborative intrusion detection method based on reputation weight and trusted distribution. This approach first employs differential privacy and secret sharing techniques to support locally trained detection models, dynamically evaluating node reliability through subjective logic. Subsequently, it combines a reputation-weighted aggregation algorithm to effectively suppress malicious attacks and enhance system stability. Finally, blockchain technology is introduced to establish a trusted distribution mechanism, ensuring secure model updates. Simulation experiments on the CBTCset dataset demonstrate that the proposed method achieves an accuracy rate of up to 99.6%, outperforming traditional average-weighted aggregation and privacy-preserving methods in terms of accuracy, F1, precision, recall and time delay.