This paper mainly studies the web log generated by the data center, with considering all data for both get and post requests, constructs the feature vector with word2vec, and builds the model with support vector machine. Based on MapReduce parallel computing model, an anomaly intrusion detection algorithm is proposed, and a security analysis platform based on Web log is constructed. The system operation results show that the platform can effectively find abnormal intrusion in the campus network, with high retrieval efficiency, and can effectively improve the operation and maintenance efficiency and abnormal troubleshooting speed.