
274 The State of the Art in Intrusion Prevention and Detection
increase in performance and accuracy because only suspicious events are transferred to the expert
system. The main disadvantage is the necessity of updating the expert system when the neural net-
work detects a new attack whose scenarios are not included in the knowledge base.
Consider the second approach (the neural network as an autonomous system); most of the known
architecture of the neural network was analyses. Below is described some of the research examples.
A recurrent neural network was used in the case when the input data depended on an answer of
the system in previous steps. ...