Research On Authentication System Of Keystroke Sequence Based On Fuzzy Clustering

Intrusion detection is one of the most important part of network security mechanism. Usually most practical Intrusion Detection Systems (IDSs) in existence only compare the audit data and net data with the attack pattern database ,and find the actions deviating from security strategy. But this Intrusion Detection (ID) technology can't find unknown attacks using pattern matching. So we utilize abnormal detection to do it. In addition, different data sources significantly influence detection results and therefore we also needs to discover the more complex and hidden attack behaviors in data with data mining methods. This paper shows the methods of detecting intrusion using abnormal detection about stroke sequences by experiments. It mainly includes:1.This paper utilizes a new soft clustering algorithm: Section Set Adaptive Fuzzy c-means. It can get the numbers of clustering. This resolves the problem of local optimality, and utilizes fuzzy section set optimizing algorithm improving convergent rate, making the time which algorithm acquire number of clustering reduce.2. The results of experiments shown that the presented algorithm can overcome some traditional problems for c-means, and may detect many unknown intrusions. In host-based anomaly detection system, the system described by paper is effective in detecting the malicious actions while maintaining a low rate of false alarms.