Intrusion Detection Approach Using Connectionist Expert System

In order to improve the detection efficiency of rule-based expert systems, an intrusion detection approach using connectionist expert system is proposed. The approach converts the AND/OR nodes into the corresponding neurons, adopts the three-layered feed forward network with full interconnection between layers, translates the feature values into the continuous values belong to the interval 0,1, shows the confidence degree about intrusion detection rules using the weight values of the neural networks and makes uncertain inference with sigmoid function. Compared with the rule-based expert system, the neural network expert system improves the inference efficiency.