Intrusion Detection System using Self-Organizing Maps

More networks are connected to the Internet every day, which increases the amount of valuable data and the number of resources that can be attacked. Some systems have been designed and developed to secure these data and prevent attacks on resources. Unfortunately, new attacks are being created everyday, which makes the design of system that could catch these attacks harder. The need is not only for preventing the attack, but also to detect such an attack, if it happens. Intrusion Detection Systems is built to accomplish this task and complement other security systems. In this paper we build an Intrusion Detection System using a well known unsupervised neural network, namely Kohonen maps. We propose two enhancements that were able to solve one of the shortcomings of the available solutions, namely high value of false positive.