Based on the problem that existing antivirus software cannot effectively respond to emerging viruses and lacks real-time detection and effective interception of viruses, a class of detection propagation models based on log feature detection of viruses is established based on the classical susceptible-exposed-infected-immune (SEIR) propagation model with the introduction of containment nodes. The experimental results show that the drop rate of infected nodes of this model is increased by 26.45%, 40.68% and 54.24% when the detection rate is taken as 0.5,0.7 and 0.9, respectively, compared with the previous SEIR model. Therefore, the detection propagation model based on the detection of log feature information of viruses can effectively contain and remove the spread of new viruses in the network, and the higher the detection rate, the better the effect.