Sequential Pattern Mining Intrusion Detection System

With the development and enrichment of the Internet application, the Network data flow increases and network security becomes more and more important. How to find out and reduce security incidents seems to be one of the important problems of Internet security. Based on the traditional sequential pattern mining algorithm, this essay improves the existing PPM prediction model and prediction algorithm. Consequently it can predict intrusion behaviors more accurate by putting into use to intrusion detection system.The main research work has the following aspects:1. The research focuses on the intrusion detection accessing characteristics and experimental verification. The main features of accessing characteristics for users' access to intrusion detection access object are uneven. Zipf's 1st law can be used to depict the high frequency accessing objects'popularity and for the low frequency objects we borrow Zipf's 2nd law to model. Accessing is a random variable, it obeys inverse Gaussian distribution. The deep research on the intrusion detection accessing characteristics provides the theoretical basis to the improvement of prediction model.2. This paper improves the existing PPM prediction model. The core of the model is based on the accessing characteristics of the PPM prediction model. And besides inherits features of traditional model, this new prediction model uses pruning technique, pruning are carried out pre-and post-pruning, this model achieves the purpose of controlling the scale of the model. Finally, the intrusion detection system log files use the data to the new PPM prediction model, using simulations to verify its validity.