B. The Generation of Network Security Situation
There are two network security situation data sources
available for knowledge discovery: one is the set of security
alert events generated from the attack simulations, the other
is the set of historical security alert events. The function of
knowledge discovery in our framework is to find out and
extract the knowledge from these set of alert events, which is
required for the correlation of security situation. Due to the
complexities of alert events generated from various types of
security situation sensors, the process is hardly to be
performed completely by manual work. In this paper, we
propose a knowledge discovery based method, which
provides the means of extracting the security situation
correlation rules through the pattern mining, analysis and
learning from the set of security alert events, and finally
generate the network security situation graph. This process is
divided into the following steps:
1) Simplification and Filtering of Security Alert Events
We found that there exists large numbers of meaningless
frequent patterns in the set of primitive alert events from
security situation sensors by examining the experiment data,
and these frequent patterns mostly relate to the problems of
system configuration or harmless access. If the process of
knowledge discovery is directly performed on such set of
primitive intrusion events, it is inevitable to generate many
types of meaningless knowledge. Therefore, it is necessary to
establish the mechanism of alert event filtering in the
foundation of D-S evidence theory, which executes the
statistical analysis based upon the confidence level of alert
events. Firstly, the distributions of various types of security
events are statistically analyzed via automatic tools;
secondly, the meaningless events are deleted by evaluating
the importance of each type of alert events based upon the
rules of simplification and filtering, which uses D-S evidence
theory as the foundation of event processing.
2) Knowledge Discovery from the Set of Security Alert
Events
In this paper, the frequent pattern and sequential pattern
discovery algorithm are adopted to obtain the security
situation knowledge from the set of security alert events. The
frequent pattern refers to the correlations among the