The stealth performance of submarines is closely related to their hull forms. In this study, an optimization method based on Deep Reinforcement Learning (DRL) was developed to design submarine hull forms, aimed at maximizing the stealth performance. The DRL optimization technique relied on the decision-making process of an agent for determining actions resulting in changes in the hull form, using stealth performance as the reward. The stealth performance of the submarine was evaluated through a Target Strength (TS) analysis model. Additionally, functional constraints of the examined hull forms were implemented in the optimization process, including geometric constraints related to the hull form and dynamic stability constraints pertaining to the hydrodynamic maneuvering characteristics. The TS of the final optimized hull form was 6.5 dB lower than that of the base model, indicating remarkable stealth performance and improved maneuverability. These results validated the effectiveness of the proposed DRL-based optimization method.