It is well-known that information security means the protection of information, and ensuring the availability, confidentiality and integrity of information. The purpose of this paper is to present an improved RBF neural network method for information evaluation. Ant colony optimization is a multi-agent approach for difficult combinatorial optimization problems, which has been applied to various NP hard problems. Here, ant colony optimization algorithm is applied to optimize the parameters of RBF neural network. In this paper, we employ “unauthorized access”, “unauthorized access to a system resource”, “data leakage”, “denial of service”, “unauthorized modification data and software”, “system crash” as the features of information security evaluation. It is indicated that the information security evaluation error of the improved RBF neural network is smaller than that of the RBF neural network. Thus, the improved RBF neural network is very suitable for information security evaluation.