The duty of securing networks is very difficult due to their size, complexity, diversity and dynamic situation. Currently applying neural networks in intrusion detection is a robust approach to ensure security in the network system. Further, neural networks are alternatives to other approaches in the area of intrusion detection. The main objective of this research is to present an adaptive, flexible and optimize neural network architecture for intrusion detection system that provides the potential to identify network activity in a robust way. The results of this work give directions to enhance security applications such as Intrusion Detection System (IDS), Intrusion Prevention System (IPS), Adaptive Security Alliance (ASA), check points and firewalls and further guide to the security implementers. © 2010 Springer-Verlag.