In response to the difficulty of quantifying the fuzziness and randomness of safety risks in highway engineering construction using traditional methods, which leads to insufficient reliability, this study proposes a risk assessment method for highway engineering construction safety reliability based on the two-dimensional cloud model and Bayesian network. An assessment system containing 4 categories and 18 indicators was constructed using the Delphi method, a fault tree model was established and transformed into a Bayesian network. Based on the risk matrix method, a two-dimensional cloud model was constructed and the comprehensive cloud of bottom events was calculated. By assessing the occurrence probability of the bottom events and the severity of their consequences, the a priori probabilities of the risk indicators were calculated. Bidirectional inference in the Bayesian network was then used to determine the probability of the top event and identify key factors.