Fuzzy Association Data Mining Techniques In Ids

With rapid development of computer network technology, the information industry and its applications have made tremendous progress. Network users face increasingly serious security issues, the need for a credible, reliable, secure and stable network platform. Hacking into a computer has been the greatest threat to security and network security. Intrusion Detection System (IDS) is the research hotspot in the field of Network Security, which plays an important role in safeguarding Network Security.Intrusion detection system is an active defense way with all kinds of network security technology. Yet in present main research work focuses on anomaly detection, it can discover the intrusion which is unknown. However, there are many flaws in it. Normal pattern of knowledge database is created difficultly, and normal pattern and anomaly pattern is not easy to explicitly divide. In view of these traits, data mining methods used to solve these problems is extremely effective. But data mining technology requires highly training data, and trains data strictly, and noise influence can make training results produce deviation and so on. Most importantly, sharp boundary problem is necessity to resolve. Therefore, fuzzy theory applies to data mining will successfully solve the problems as it mentions above.Firstly Fuzzy C-means clustering algorithm is improved in the paper. There are three main points in it. Through amending membership degree reduces isolated points which influence clustering centers. Through fuzzy cut-off set quickens convergence. Through analyzing fuzzy clustering validity can reach overall superior. The purpose that using fuzzy C-means clustering is to make fully use of its clustering results as the input of fuzzy association rules, because data which is membership degree in the fuzzy association rules is designated by an expert or by experience, even so it lacks objectivity. Therefore, applying this method will can enhance efficiency and adaptive. Secondly an improved algorithm using hashing tables on mining fuzzy association rules is proposed in the paper, and equivalence classes are introduced to search frequent item sets quickly. With this algorithm the usual practice of repeatedly database scanning can be avoided. Its efficiency is showed with a typical use on Intrusion Detection System (IDS). So it greatly improved detection speed and the efficiency of the fuzzy intrusion detection systems.