The invention discloses a
network security accident classification and predicting method and a
system, which are used for solving the problems of the prior art on lacking the capability of timely finding the characteristics of
attack behaviors and accurately classifying the
attack behaviors. The method comprises the following steps: S1. acquiring
web access log of users in a whole network and http
metadata in full-flow log; S2. segmenting the
web access log and the url of the http
metadata, and matching with a network attach illegal character feature
library; S3. constructing a word vector and a document vector of the segmented url by utilizing word2vector; and S4. inputting the document vector as a feature and classifying the
attack behavior by adopting a naive bayes model. The real-time monitoring of key points can be realized, the abnormal behavior carrying mainstream attack feature can be found by means of
machine learning, the classification efficiency of
network attack behaviors can be improved, and the
time cost of manual check can be lowered, continuously changed attack behaviors can be adapted, and the classification detection accuracy can be enhanced, thus providing guarantee for
network security.