IDS Adaptation for an Efficient Detection in High-Speed Networks

Intrusion Detection Systems are essential in a network security solution. However, with the significant development of network technologies, the current IDS architecture does not support high-speed communications. Therefore, improving the performance of IDS is a major concern for researchers. In this paper, we present a model of intrusion detection based on the classification of network connections. Our approach is based on the principle of an intelligent loss. We propose a classification model based on the principle that a connection is either malicious or not. In the first case, the connection must be handled by the IDS; otherwise we can ignore it. This method reduces significantly the network flow sent to the IDS with a tolerance of an error threshold. This threshold can be adjusted by the updating process of the classification model.