In order to solve the problems of poor security, low precision and recall rate in scientific research audit system, a security vulnerability identification method of scientific research audit system based on distributed machine learning algorithm is designed. Collecting the user data of the scientific research audit system, clustering the user node data, introducing the concept of the k-nearest neighbor (KNN) algorithm to establish a network distributed structure model of the scientific research audit system, and combining and classifying the security vulnerability characteristics with representativeness and diversity. Based on distributed machine learning algorithm, security vulnerability identification is carried out in practical application. Two traditional security vulnerability identification methods are compared. The results show that the method can identify different types of security vulnerabilities, and the accuracy, precision and recall are improved.