Neural Networks in Vulnerability Assessment

According to the statistics reported by the Internet information center over the years, the number of hacker attacks on computer users worldwide increases by at least 10% on average every year. Network vulnerability analysis is a very complex work, and the correlation among vulnerabilities, the correlation between network hosts, the dynamic nature of network services and the complexity of network connections are worthy of attention. With the increasing attention paid to computer network security, computer network vulnerability assessment has important research and application value.

Due to its inherent super adaptability and learning ability, neural network has been widely studied and applied in many artificial intelligence fields and has solved many information processing problems that are difficult to be solved by other traditional artificial intelligence methods and technologies. Because of the unique ability of nonlinear adaptive information processing, neural network overcomes the shortcomings of many traditional artificial intelligence information processing methods in pattern recognition, voice information recognition, unstructured information processing and other visual functions, so that it has been successfully applied in many fields of artificial intelligence. The close combination of neural network and other traditional information processing methods of artificial intelligence will greatly promote the continuous innovation and development of related technologies such as traditional artificial intelligence and distributed information processing.

Several factors must be considered when designing large interconnections. Optimal design is important to achieve good performance and reduce construction and maintenance costs. Real communication networks are prone to network component failures. There are failures between nodes and connections, and network stability is desirable when a limited number of failures do not disrupt the entire system. The vulnerability of network topology is an important aspect of computer network design. Aysun Aytac proposed various methods to quantify network vulnerability and derived network reliability formulas using a large number of graph theory parameters. Based on the control concept in graph theory and the strong and weak control number of transformation graph Gxy+, this paper studies the vulnerability of the interconnection network to the failure of a single node and measures the vulnerability of the network. Attack chart is an effective method to solve many problems in computer network security management. After a vulnerability scanner is used to identify a single vulnerability, the attack chart can relate a single vulnerability to the likelihood of an attack and then analyze and predict which privileges an attacker can gain through a multistep attack (in which multiple vulnerabilities are exploited in turn). Teodor Sommestad tested the practical application of this analytical method. The attack graph tool MulVAL obtained information from the vulnerability scanner Nexpose and network topology information from eight virtual organizations containing 199 machines. Two groups of attackers attempted to infiltrate these networks over a period of 2 days and reported which machines they had damaged and which attack paths they were trying to use. Security metrics are powerful tools for organizations to understand the effectiveness of protecting computer networks. However, most of these measurement techniques are not sufficient to help companies make informed risk management decisions. Abraham proposes a stochastic security framework that allows for quantitative measures of security by considering vulnerability-related dynamic attributes over time. Abraham's model is that existing research in attack graph analysis does not take into account the timing of vulnerabilities, vulnerabilities and patch availability, etc., which can be interlinked based on how the entire network is affected by vulnerabilities and leverage compromise systems. In order to more realistically describe the changes of the security state of the network over time, an inhomogeneous model is proposed, which contains a time-dependent covariable.

In this paper, based on neural network are studied under the study of computer network vulnerability analysis method and the model of network vulnerability index calculation, on the basis of clear vulnerability index and related concepts, emphatically discusses the vulnerability analysis method and based on the index system of vulnerability and vulnerability database index calculation model, aims to provide a kind of multidimensional network security status display, macro, so that the relevant enterprises and departments in a timely manner to master the Internet network security macro situation.