In response to the problem of multiple types of intrusion data in information emergency networks, a multi-layer anti intrusion algorithm for information emergency networks is proposed, which combines active and passive methods. This strategy combines the technological advantages of active defense and passive detection, and constructs a multi-level and all-round protection system. Passive detection adopts a network coding method based on information theory security, which encodes and decodes data packets through linear operations to ensure the confidentiality of information during transmission; Active defense uses support vector machine algorithm to achieve intrusion detection. By introducing kernel functions to map data to high-dimensional space, the optimal separation hyperplane is found in the high-dimensional space and mapped back to the original space, thus achieving non-linear classification of intrusion data. The experimental results show that this method can accurately detect various types of network intrusion data, and it also has the ability of self-learning and optimization, which can continuously adapt to changes in the network environment and improve the protection effect.