With the increasing cyber security threats, malware attacks pose a serious challenge to the security of information systems. This study proposes a malware behavior detection and protection method based on Bayesian inference model. Accurate detection and protection against malware behavior is achieved by constructing a state space model with Markov chain process, combined with Bayes theorem. Experimental results show that the method in this paper has high accuracy in malware attack detection. In the simulation experiments, using the Markov chain algorithm for parameter estimation, the relative error of the final malware location parameter is 0.0366%. Meanwhile, in the algorithm error adaptation analysis, the parameter estimation interval gradually increases with the increase of the total error standard deviation, but through data cleaning, the error adaptation is significantly improved and the parameter estimation interval is narrowed. Experiments show that the method can effectively improve the recognition accuracy of malware attacks, especially in the case of concurrent attacks, the decision rate and detection rate are better than the traditional method. The study proves that the proposed Bayesian inference-based malware behavior detection and protection method has high accuracy and robustness, and can effectively identify and defend against malware attacks.