Regression Based Network Security Scenario Prediction Model

Abstract

Analysing issues in which one or more independent variables predict an outcome may be done using logistic regression. A binary or dichotomous dependent variable is used to quantify the result, which only comprises data coded as 1 (True, Success, etc.) or 0 (False, Failure, etc). Logistic regression is used to identify the best model to represent the connection between a dependent variable (outcome or response variable) and a collection of independent (predictor or explanatory) variables. Biomedical applications of LR (Linear Regression) include cancer detection, survival prediction, and more. This is a popular and well-established data analysis approach in statistics and biomedicine. It is also recommended to compare data mining approaches with logistic regression when mining clinical data. In this research, various ideas about network security assessment are described and compared,, and a technique for evaluating the current state of network security using a virtual Honeynet is proposed. A binary linear regression model is developed based on the correlations between Honeynet active, host active of computer networks, and IP active when network intrusion occurs. Prototype systems for data gathering and regression fitting are built, proving the validity of regression prediction models.