基于PSO-LightGBM的网络资产脆弱性评估模型

Vulnerability assessment model of network assetsbased on PSO-LightGBM

摘要:

随着网络空间资产探测技术的不断发展,越来越多的资产脆弱面暴露在公众面前,在一定程度上增加了网络资产的安全风险。对网络资产进行脆弱性评估,可以及时发现脆弱性较强的高危资产,在安全事件未发生时主动对脆弱的网络资产进行保护和修复,从而有效降低网络安全事件发生的概率。现有研究主要集中在网络资产漏洞评估及网络系统脆弱性评估上,对网络资产脆弱性评估方法的研究还比较匮乏。为了更好地保护网络资产安全,提出了一种基于粒子群优化算法轻型梯度提升机(particle swarm optimization-light gradientboosting machine, PSO-LightGBM)的网络资产脆弱性评估模型。首先,依据行业标准和专家经验,提出针对网络资产脆弱性的评估指标体系,并根据从网络中爬取的网络资产数据,经预处理后构建了具有12个属性特征、11类标签值的网络资产脆弱性评估数据集;其次,将PSO 算法与LightGBM 模型相结合,利用机器学习方法实现网络资产脆弱性的自动化评估;最后,通过实验对比了几种机器学习模型在数据集上的表现,结果表明,基于PSO-LightGBM的网络资产脆弱性评估模型的评估准确率可以达到91.24%,充分验证了该模型的有效性。

Abstract:

With the development of cyberspace assets detection technology, more and morevulnerable assets are exposed to the public, which will increase the security risk of cyber assetsto a certain extent. Vulnerability assessment of network assets can help people discovervulnerable and high-risk assets in time, and proactively protect and repair vulnerable networkassets when security events do not occur, which can effectively reduce the probability of networksecurity events. The existing researches mainly focus on the vulnerability assessmentsof the network assets and the network system, but rarely on the vulnerability assessment methods. In order to better protect the security of network assets, a vulnerability assessmentmodel of network assets based on particle swarm optimization algorithm-light gradient boostingmachine (PSO-LightGBM) was proposed. First, according to the industry standards andexpert experiences, an evaluation index system for vulnerability of network assets was proposed.On the basis of the network asset data crawled from the network, a network asset vulnerabilityassessment data set with 12 attribute characteristics and 11 types of label valueswas constructed after pretreatment. Then, light gradient boosting machine(LightGBM)model was combined with particle swarm optimization (PSO) algorithm to realize automaticvulnerability assessment of network assets by machine learning method. Finally, the effectivenessof the network asset vulnerability assessment model based on PSO-LightGBM wasverified by comparing the performance of several machine learning models on the data set.The experimental results show that this model can accurately assess the vulnerability of networkassets, with an accuracy of 91.24%.