Abstract: When designing the waterproofing and drainage system for an subsea tunnel, it is crucial to have a clear understanding of the surrounding rock's inherent water-blocking capabilities to achieve active control over the drainage volume. This paper first introduces the concept of the surrounding rock impermeability, defined as the ability of the rock’s ability to resist water infiltration into the tunnel. A predictive formula for water inflow is derived, taking into account nonlinear seepage conditions in fractured rock masses. The formula considers several factors that influence impermeability, including engineering geological conditions, hydraulic connectivity within the rock, and the size of the tunnel. On this basis, the study performs a statistical analysis of water inflow data from 52 typical subsea tunnel sections, particularly those in water-rich environments, to identify key factors influencing the impermeability of surrounding rock. These factors include rock cover thickness, hydraulic head, uniaxial saturated compressive strength of the rock, and volumetric joint count. These factors are used as indicators to establish a classification standard for the rock impermeability. To enhance the classification process, machine learning techniques are employed. The bisection method and information gain ratio from the training dataset are used to analyze the data. A decision tree model capable of handling continuous-valued attributes is established. This model allows for the classification of surrounding rock impermeability based on the relevant rock parameters, thus enabling a more automated and data-driven approach to impermeability classification. Finally, the model is applied to the drill-and-blast section of the Qingdao-Jiaozhou Bay Second Subsea Tunnel, verifying the rationality and feasibility of the proposed anti-seepage classification method. The research findings provide a theoretical basis for determining drainage control standards in subsea tunnels. Compared to traditional rock mass classification methods, the anti-seepage classification method comprehensively considers the conditions of the surrounding rock and its seepage mechanical response, leading to a more scientific and reasonable approach to waterproofing design and zoned drainage strategies.