When Loitering Munitions (LMs) perform adversarial tasks, they face challenges such as unknown interception strategies of Non-Cooperative Targets (NCTs), measurement noise, strong system coupling, and non-convexity. To address these challenges, this paper proposes an interception evasion algorithm for LMs based on optimized risk budget allocation. The algorithm adopts the Alternating Direction Method of Multipliers (ADMM) to decompose the original coupled non-convex problem into a convex control optimization subproblem and a risk budget allocation subproblem with an analytical solution, and solves the two subproblems alternately via iteration. In particular, the risk budget equation establishes dynamic associations among estimation uncertainty, interception situation, and risk budget, and enables the algorithm to effectively deal with NCTs with heterogeneous maneuvering capabilities. Simulation results demonstrate that the proposed algorithm ensures that the LM cluster achieves reliable and safe evasion performance under three typical NCT interception strategies.