
350 The State of the Art in Intrusion Prevention and Detection
14.5.3.3 Bayesian Systems
Bayesian networks model probabilistic relationships between variables of interest and are very simi-
lar to neural networks. Here, connections represent conditional dependencies, and nodes that are not
connected to each other represent variables that are conditionally independent, regularly described
as a DAG (directed acyclic graph) [28] as in Figure 14.10. In a DAG, each node represents a domain
variable, and each edge between nodes indicates a dependency, usually based on probabilities. Thus
the probability of the event occurring is based on the evidence ...