In this paper, we analyzed the use of neural network for packet filtering. The neural network system was designed in eight ways with input to the neural network in the form of either access rules or optimized access rules or binary form of access rules or representing wildcards as 0 & 255 or combination of them. These trained neural networks were analyzed for their correctness and the performance aspects such as training time using test data. In order to further improve the security, the data related to the local usage of the network were also used to train the network. An example of implementing these trained systems in active networks packet filtering was presented.