No abstract

No abstract

No abstract

In this paper, we propose a hybrid intelligent intrusion detection system to recognize novel attacks. Current works in intrusion detection solve the anomaly detection and the misuse detection. The misuse detection cannot recognize the new types of intrusions; while the abnormal detection also suffers from the false alarms. The mechanism to detect new forms of attacks in the systems will be the most important issue for intrusion detection For this purpose, we apply the neural network approach to learn the attack definitions and the !~IZZY inference approach to describe the relations of attack properties for recognition This study concentrates the focus on detecting distributed denial of service attacks to develop this system. Experiment results will verify he performance of the model

Conformal mappings are incorporated into the self-organization model to represent images harmonically. This network is used to partition an image into quadrilateral regions, where each region contains similar features. We then map each region to a corresponding square region to unify information representation and facilitate computations. This mapping is constructed to preserve spatial information while complying with the conformal property of the network. An approximated image in each square region provides us with an effective representation of the image in both modeling and compression applications. This approach has been particularly developed for large continues images.